Rendered at 16:11:27 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
nonethewiser 1 days ago [-]
Please read the article. Comments here are wildly misinformed.
This is a lack of linux driver support for firmware designed for windows laptops.
This is a problem on the MSI Prestige laptop using the RTX Spark platform. Lets assume it also applies to the Asus and Dell machines using RTX spark.
RTX Spark is a new platform and all of these laptops ship with Windows. You cant get Linux on any of these machines from the factory. There is also precedent for NVIDIA shipping linux support well after Windows (2 years in the case of Dynanmic Boost, a similar power sharing feature).
Zak 20 minutes ago [-]
This isn't just a matter of lack of a good driver because they didn't bother making one for an OS they didn't intend to support. Encrypting the firmware's power table serves no purpose other than frustrating any third-party effort to develop a good driver.
It is not obvious to me why they would want to do that.
callamdelaney 21 hours ago [-]
It doesn’t matter if its designed for windows, it should work as well on Linux.
alex43578 20 hours ago [-]
It’s not the OEMs job to ensure compatibility with every OS. They shouldn’t have to ensure it works well with Windows XP, for example.
herpderperator 20 hours ago [-]
Windows XP went EOL on April 8, 2014. "Linux", assuming you mean the latest Linux kernel, is not.
alex43578 19 hours ago [-]
Do they also need to support MacOS, Chrome OS, and FreeBSD? They made these laptops to work with Windows. It’s not on them to ensure compatibility with anything more than they were marketed for.
bigyabai 18 hours ago [-]
> Do they also need to support MacOS, Chrome OS, and FreeBSD?
They basically do. ChromeOS runs a Linux kernel, so the userland has everything it needs to run CUDA. FreeBSD is a first-class target for Nvidia drivers, and macOS used to be supported too but stopped signing Nvidia code. Nevertheless, AArch64 is still supported for CUDA et. al..
It is up to Nvidia to support whatever platforms they want users to be on. In their case, it means supporting many OSes.
alex43578 17 hours ago [-]
The article says "A real fix depends on Nvidia, specifically on porting the platform controller’s power management logic into its Linux driver, which would effectively turn the laptop into a portable version of the DGX Spark."
My point is that's not Nvidia's responsibility. The article itself admits the laptop "ships with Windows on ARM", so support for another OS shouldn't be expected, even if NVIDIA supports other platforms and other product lines differently.
bigyabai 16 hours ago [-]
These aren't "other product lines" though. They are RTX and Spark SOC products that do support Linux as a first-class citizen with all of their hardware. The RTX Spark is the outlier, and the breakage is not on the GPU product, but the SOC devicetree drivers. This is only a problem because Nvidia avoided standard Windows power management a-la ACPI/UEFI, and has to port the remainder to Linux by hand. Non-laptop products like the DGX Spark support this via UEFI natively.
There are OEMs that refuse this responsibility. If that's the vindication you wanted, there you go. This specific issue is not a catastrophic oversight though.
alala-lala-lon 18 hours ago [-]
[dead]
mrheosuper 14 hours ago [-]
Last time i check, Both HaikuOS and React OS is still being actively developed
incompatible 15 hours ago [-]
They should at least be releasing sufficient hardware specs so that people can use the hardware they've purchased how they like.
bayindirh 6 hours ago [-]
> This is a lack of linux driver support for firmware designed for windows laptops.
Personally, I don't care. If you don't support open drivers, you have to provide first class drivers on Linux on day 1, because NVIDIA's majority of business is running on Linux already.
On top of that, they claim to have the best Linux support, and claim to have best drivers in their class.
So, it's better to act like that and provide support. If any other company has done that, they have been incinerated with the flame of internet's hate thrice already.
yunnpp 21 hours ago [-]
You are asking to read the article in a vacuum and ignore the context in which the driver support is missing. That doesn't make for very good reading comprehension.
> There is also precedent for NVIDIA shipping linux support well after Windows
There is also a precedent for NVIDIA nerfing drivers on Linux and pulling anti-competitive shenanigans against competition, so it's only reasonable that people look at the "missing" driver support with suspicion.
keybrd-intrrpt 18 hours ago [-]
And the DGX Spark and DGX Station both are Ubuntu based linux.
How does your conspiracy theory hold up to that?
GB10 and N1X are both Blackwell cousins. So they've had consumer-level Linux machines for a while now.
This was simply a partnership with Microsoft, so of course development of drivers and firmware focused heavily on Windows. Especially since I'm sure Windows required a LOT of updating to run nicely on ARM as well as the N1X unified memory model.
yunnpp 17 hours ago [-]
They have data centers to sell, and it's not out of this world for Microsoft to have requested a nerf as part of their "partnership".
You may as well be right, I am speculating and this theory might be baseless (I don't think it is.) But GP's comment is wrong to ask people to simply read the article at face value. That is bad reading.
qwerpy 22 hours ago [-]
Unfortunately HN commenters as a group are not any more intelligent or resilient to emotionally-appealing misinformation than the general public, contrary to what we'd like to believe. We see a headline that confirms our biases, we turn off our brains, and then we make in-group noises to show that we belong. Pretty much the same as religious groups and rituals that we turn our noses up at.
callamdelaney 21 hours ago [-]
Is it misinformation that the driver is slower on linux? There’s no such thing in my book as ‘windows’ hardware. It should work as well everywhere on day one, or we should have the ability to make it do so.
keybrd-intrrpt 18 hours ago [-]
> There’s no such thing in my book as ‘windows’ hardware.
Then you misunderstand how hardware, software, firmware, drivers, and operating systems have worked for the last few decades.
vondur 16 hours ago [-]
We used to have things called WinModems back in the day. Skipped using hardware modems and ran it on software that only worked under windows.
ThrowawayB7 10 hours ago [-]
> "There’s no such thing in my book as ‘windows’ hardware."
Good luck with that. A PC is more or less defined by specifications laid out in Microsoft's Windows Hardware Compatibility program and tested by their Windows Hardware Lab Kit. Before that, it was defined by IBM.
I've heard that ARM has had a standard for ARM-based servers, their Server Base System Architecture, but I've never seen any compliant hardware in the wild.
m3kw9 23 hours ago [-]
This is the same reason misinformation works so well on these people they have to stuff and they don’t critically think if there are other reasons. They’d write more headlines and others will act the same, like wild fire.
jleask 1 days ago [-]
Nvidia doing what Nvidia has always done; proprietary closed drivers. Weird choice to cryptographically protect the power settings. Key problem is that we've allowed a company that's always been this way to form a monopoly. Other brands exist, perhaps it is now worth the effort to migrate and rebuild elsewhere.
r3trohack3r 1 days ago [-]
> Key problem is that we've allowed a company that's always been this way to form a monopoly.
I don’t think anyone “allowed” them - quite the opposite. They’ve made generational bets that ended up being right, bets that IIRC the industry did not agree with for a long time. Them pursuing CUDA resulted in a draw down of something crazy like 75% of their valuation. They bet the farm multiple times to get to the point they are - and it’s only “obvious” in hindsight.
If you want folks to make generational bets as a matter of course (which I think most folks do), centralizing the downside of those bets and socializing the upside of those bets is not a great strategy. Almost as bad as socializing the downside.
Competition will eventually close that gap. We’re only a few years into this thing.
thrance 1 days ago [-]
Monopolies don't naturally decay, competition can't close the gap against an entrenched titan that can either buy or asphyxiate them.
krapht 23 hours ago [-]
I know NVIDIA is dominating the market, but it's not like AMD/Intel/Google don't have credible alternatives. Hipify/OneAPI/MLIR(Jax etc) can be used today, it's just not as ergonomic as CUDA.
logancbrown 22 hours ago [-]
You could have done maybe 20 seconds of research to find what you posted is not correct "Monopolies don't naturally decay".
Have you ever heard of U.S. Steel, IBM, Kodak, and Western Union?
thrance 22 hours ago [-]
These monopolies were defeated (or weakened) by new technologies they didn't adopt early enough, not by market forces like competition.
logancbrown 20 hours ago [-]
New technologies are indeed how natural competition market forces primarily work.
thrance 16 hours ago [-]
Hmm... I kind of disagree with that. Technological breakthroughs are not "natural". We shouldn't have to bet on someone discovering something that is able to make NVIDIA obsolete.
afarah1 1 days ago [-]
New generations such as Blackwell have open source kernel modules, there is not even a proprietary option. I thought there was a move in this direction?
Nvidia enabled that by offloading stuff that used to be in the proprietary driver onto the still proprietary GPU firmware. Making it, if anything, less open source as it's now harder to inspect what it's doing at runtime since it's now out of reach of even the kernel
maxhille 1 days ago [-]
AFAIK the open source kernel driver pushes a signed proprietary 500Mb firmware blob into the GPU.
NekkoDroid 1 days ago [-]
The total linux-firmware-nvidia package for Archlinux uncompressed (outer, the firmware blobs are also compressed on top of that) is ~160MB (Mb would be Megabit), with the biggest individual compressed blob being ~50MB.
maxhille 1 days ago [-]
You are right the firmware packages are smaller. I mixed it up with the VRAM the GSP uses after being loaded.
Point was about the kernel drivers being open-source enabling hackability - they don't, the card hardware is just as closed as before.
mrheosuper 14 hours ago [-]
50MB is a lot of code. Linux kernel in minimal config will use significantly less than that.
NekkoDroid 6 hours ago [-]
Never wanted to imply it wasn't. AMDGPUs firmware package is a total of 27MB and the biggest single blob is 1,5MB.
wqtz 1 days ago [-]
> we've allowed a company
When did "we" have a say in this? I don't get these types of comments from the "open-source as a community."
Nvidia has been doing this for decades, and there are two possible arguments:
1. They are successful despite it.
2. They are successful because of it.
There is no mythical collective "we" representing the open-source community in Nvidia's supply chain of success. It's Nvidia and its consumers.
tancop 1 days ago [-]
It's "we" as a society. Not pushing our governments to pass laws that could make Nvidia open up its hardware. Of course it's not your fault or mine, the problem is money in politics and uneducated voters.
ion_trapper 19 hours ago [-]
As we speak, there are numerous projects being worked on that you have no idea will be immensely useful to you in 10 years, and the people who are working on these projects are more likely than not motivated by the potential of future profit. 10 years from now, when you have recognized the value of these projects, you will take them for granted and demand you get them for free. CUDA was one of those projects 10 years ago. “We” can choose to start the precedent of scavanging CUDA today, but the price will be that those projects 10 years down the road won’t materialize.
NoidFonsense 1 days ago [-]
how are they going to sell you a subscripton to overclock your GPU otherwise?
jleask 1 days ago [-]
Sad but probably close to the truth. I'm old enough to remember when companies were able to make money without actively robbing their customers
Orygin 1 days ago [-]
They always have been robbing their customers. They have sold locked down GPU for ever
bigyabai 22 hours ago [-]
And people keep coming back. Myself included; there's no similar GPU products from AMD, Apple or Intel.
Nvidia won because their architecture is better. If you do a significant amount of work that can be CUDA-accelerated, you're kinda wasting your time not owning a CUDA-capable card. It's like being a video editor that software-encodes their projects because they don't want to get an accelerator card. It's your choice, but software-acceleration is a nightmare compared to how quick and efficient hardware-acceleration is. Competitors will all use the accelerator card and ship faster than you can.
Apple, AMD and Intel have all earned their current position by abandoning Khronos and refusing to build their own GPGPU architecture. If the entire industry refuses to invest in complex technology just because it's expensive, then they don't deserve to make money.
jack_pp 1 days ago [-]
they could make money that way right now but they figured out they can make even more money by doing this. but jokes on them, people will figure it out, unlock / jailbreak and they just lose good will long term
dist-epoch 1 days ago [-]
It's not weird at all, they know people will buy these laptops, put linux on them and stash them in data-centers, so they are trying to prevent this.
Roark66 1 days ago [-]
Not those laptops.
I was genuinely excited about this device until I heard the specs. Still under 300GB/s memory bandwidth when a gpu like rtx3090 from 6 years ago had 900+
This is really bread crumbing local inference users.
We should be measuring our memory bandwidth in TB by now. The only company that seems to be serious about this is Apple. (the m5 ultra studio etc).
jasomill 1 days ago [-]
If this were cost-effective, why wouldn't they just buy DGX Sparks, with similar performance at similar price points and a more convenient form factor?
Or RTX Spark and run their workloads under WSL, which as the article notes, has minimal overhead relative to running Linux directly?
This sounds more like a bug than a feature, especially given that Linux on battery runs at full clock speed and downclocks when plugged in, and Windows does the opposite, suggesting they could at least mitigate the problem with no changes to their software stack by inverting a conditional in the firmware and adding a Linux kernel parameter to control memory allocation as apparently already exists for Windows.
kllrnohj 1 days ago [-]
that seems really hard to believe? If you wanted this running Linux in a data center you'd have just been buying the cheaper and yet also more powerful DGX Sparks.
The only value proposition of RTX Spark at all is the laptop form factor
Roark66 1 days ago [-]
No sane perse would buy dgx spark for datacenter. If you need Blackwell for the native nvfp4 you'd be better buying pallets of rtx5060 ti 16GB. Or if you don't, pallets of used rtx3090s.
In fact I'm pretty sure it's happening. Why? Because of constantly growing prices of those things.
kllrnohj 1 days ago [-]
Sure, but if you think it's insane to buy a DGX Spark for a datacenter, then it's equally insane to buy an RTX Spark (which is the exact same SoC just with less networking and a screen) for the datacenter.
dist-epoch 1 days ago [-]
> NVIDIA says that the new DGX Spark 64GB is starting at $4,999
> NVIDIA has now confirmed an official starting price of $2,399 for RTX Spark laptops
Nvidia's MSRP is full of shit and has been for a long time. You see it with desktop GPUs all the time, even before the AI boom, where Nvidia would make up an MSRP but not offer a card at that price themselves, and none of the AIBs would, either, except for some token amount that rapidly sold out and were never restoked.
Hell there doesn't seem to be an RTX Spark that's sub $3,300 when considering the 32GB models even. $2,399 is pure fiction.
jasomill 1 days ago [-]
The minimum configuration appears to be 24GB / 18 core, which at least one vendor, Microsoft, is selling for $2,599.99, which is at least closer to $2,399 than $3,000+.
jasomill 1 days ago [-]
If datacenters aren't lining up to buy 64GB DGX Sparks for $5,000, what makes you think they'd be interested in 24GB RTX Sparks with fewer cores and less attractive form factors for almost half that?
And if there's a market for a DGX Spark configuration with a higher compute-to-RAM ratio, why isn't Nvidia offering it?
My best guess is that the entry-level Sparks are mostly intended for people who would otherwise be in the market for a traditional high-end laptop, who might be enticed to spend a few hundred more for at least the promise of local AI, while the higher-end models are intended for more serious AI development.
With that said, it's not hard to imagine there isn't nearly as large a built-in market for high-end SFF Linux PCs as there is for high-end Windows laptops.
aurareturn 1 days ago [-]
Yes, exactly this.
Nvidia does something to prevent abuse so that regular people who want the laptops can buy one. Nvidia gets knocked on.
kllrnohj 1 days ago [-]
Regular people aren't buying a $4,500 laptop lol wtf are you talking about?
This is all but certainly a bug in their linux driver and nothing else. It sure as hell isn't some "anti abuse" to enable "regular people" to buy one.
fxwin 1 days ago [-]
> A hands-on look at dual-booting Linux on the Nvidia RTX Spark laptop, the GPU clock throttling bug found, and why WSL still wins.
Man it really is exhausting to read a title that sparks my interest, just to be greeted with the most basic Claude-isms in the subheading.
geoffhill 1 days ago [-]
[flagged]
fxwin 1 days ago [-]
What an odd response, but if by "foreign influence campaign" you mean "masses of badly written blog posts that i regret wasting my time reading" then sure
officeplant 22 hours ago [-]
Please be aware that you have been the target of national AI influence campaigns. Your comment history makes it clear they have had a strong negative effect on you.
mjhay 1 days ago [-]
We must stop the foreign anti-ai conspiracies from sapping and impurifying our precious bodily fluids
aa-jv 1 days ago [-]
> sparks my interest
I see what you did there. ;)
teiferer 1 days ago [-]
Enumerating 3 things really ..sparks a comment nowadays? Doing that as a rhetorical device has been a thing for generations and of course LLMs mimic it because it is everywhere in their training data.
I just heard a quick J. Talarico speech at some country concert. In the 10 or so sentences he spoke, he had 3 or 4 instances of doing this "Claudism" enumeration. Are you saying Texas just elected Claude for Senate?
Maybe we should stop with such superficial judgements. It's the modern equivalent of not liking the speaker's tie. (Doesn't mean the speaker's speech is good, but it doesn't mean it's bad either.)
fxwin 1 days ago [-]
There are more Claude-isms in there than just rule of three (Hence -isms and not -ism), and while I'd generally agree that we generally shouldn't be superficial with our judgements, this particular heuristic has served me pretty well so far. I don't know what Talarico speech you are talking about, so i won't comment on that (But I'm not sure if political speeches are the best rebuttal to "I think obviously AI generated text is an indicator of low signal-to-noise ratio").
Edit re "equivalent to not liking the speaker's tie":
This is not equivalent at all, this is a written blog post and I'm put off by the way it is written, not by say the URL or their chosen font. It's more akin to a speaker dropping an inappropriate joke at the start of their talk and leaving because of that
lemagedurage 1 days ago [-]
I think it's the combination of the profound writing style with the underwhelming substance that bothers people.
quietbritishjim 1 days ago [-]
"Why XYZ still wins" is also a very common pattern for Claude. Granted it's also found in pre-Claude clickbait titles.
IshKebab 1 days ago [-]
Speeches have always over-used these writing techniques. The problem is Claude uses them all the time, even when not writing speeches.
K0balt 1 days ago [-]
Wtf. Are we actually seeing a push towards being 100percent dependent on the cloud?
Sovereign AI has never been more important.
Controlling AI is so much more powerful and impactful than controlling search or social algorithms.
Centralized AI a 10x on social control over everything we have seen before, and there is no way that a free and open society survives contact with centralised, high capability AI if the system is manipulated towards guiding outcomes. It’s essentially mind control, as people will offload an even greater share of their cognition to AI than they do with search and social.
Build and maintain sovereign AI while you still can.
nonethewiser 1 days ago [-]
>Wtf. Are we actually seeing a push towards being 100percent dependent on the cloud?
This is a linux driver problem. On laptops designed for windows.
Read the article and stop sensationalizing.
This is a problem on the MSI Prestige laptop using the RTX Spark platform and Linux. Lets assume it also applies to the Asus and Dell machines using RTX spark and Linux.
RTX Spark is a new platform and all of these laptops ship with Windows. You cant get Linux on any of these machines from the factory. NVIDIA may even ship drivers for linux later as they did with Dynamic Boost.
Kindly remove this from your list of "Evidence that NVIDIA is Colluding to Centralize Inference" and everyone else take note of this confirmation bias in action.
advael 22 hours ago [-]
I dunno about colluding to centralize inference, but "we encrypted the power firmware and only gave microsoft the keys" certainly reads as colluding to keep people on windows artificially
prcrstntr 21 hours ago [-]
Nvidia probably encrypts firmware to protect IP rather than to block Linux users; maintaining open drivers across platforms is expensive, and they’ve historically prioritized Linux for datacenter
advael 20 hours ago [-]
Protecting IP is an excuse given for all manner of computer-aided fuckery, certainly
nomel 20 hours ago [-]
Power per compute is ludicrously important to data centers. This isn't some small IP, it's a potential market differentiator.
edit: even more important for mobile computers
advael 19 hours ago [-]
Yes, it seems a lot of companies can make a lot of money from fuckery like controlling what their customers can do with the devices they sell and deliberately sabotaging those devices for some customers via this control. What's your point? Am I supposed to support any practice a company can find a way to make money off of? Generally speaking I'm more against fuckery that's profitable for a company than fuckery that isn't, because the latter is at least likely to stop naturally
nomel 19 hours ago [-]
> What's your point?
That them trying to protect IP around something very profitable isn't crazy. That not being fully supported in Linux, when the product was literally designed around Windows, also isn't crazy or really evidence of anything, including evil (see the whole fucking history of Linux compatibility of made-for-Windows hardware).
Maybe let them cook.
advael 18 hours ago [-]
In general, the history of linux compatibility for nvidia specifically has been the research and open-source communities dragging them kicking and screaming into compliance while they do significant work to prevent compatibility via fuckery. I've been doing work impacted by this continuously since before GPGPU programming became popular. I'm not a beneficiary of nvidia's deals with microsoft or their stock price in general, so I see no reason I should "let them cook"
rcxdude 16 hours ago [-]
This kind of thing is far better at locking down the average user than preventing competitors from reverse engineering what they're doing. (much like DRM is pretty good at stopping your average consumer from doing what they want with their media but pretty bad at stopping the pirates getting access to the same content).
Peaches4Rent 8 hours ago [-]
Boo hoo. The biggest, most leather jacketest company in the world decides it can't maintain a Linux firmware. The same thing it's entire server system depends on. Omg why would that be??
It obviously can't be because of ulterior motives!
vee-kay 20 hours ago [-]
[dead]
saidnooneever 19 hours ago [-]
not being able to read something which is encrypted is not a driver problem. do u expect the driver to crack the encryption? or guess the key?
nvidia has always been crap like this. Look at amd and intel, they let u write proper drivers (even if its real difficult to do so). No encrypted blobs or voodoo needed.
the nvidia drivers u need to use a card on Linux are unintelligible binary blobs. thats sad as it is. this is worse. they might give some shitty blobs at some point, after they made enough ppl give up on it and switch to give them and their partners their money, souls and firstborn
Matl 22 hours ago [-]
> This is a linux driver problem. On laptops designed for windows.
No. It's 'the NVidia way or the highway' problem. NVidia is so full of themselves they refuse to work under anyone's rules (i.e. the Linux kernel where they're constantly trying to work around them).
They're egomaniacs who are so full of themselves, riding high on the AI bubble, they think they're above everyone else in the ecosystem, while utilising all kinds of other people's work for their own benefit.
Look at the way they develop their 'open' Linux drivers where it's just a shim that loads a huge binary blob off the GPU rather than a proper driver like amdgpu has.
Meanwhile they wouldn't be a $5T company without Linux and the wider FLOSS ecosystem.
It's kind of like when Secure Boot debuted and the plan was to allow only Windows with it enabled, then no doubt someone would claim the fact that only Windows can boot with it enabled as a 'Linux driver problem'.
shevy-java 21 hours ago [-]
Where do you see "sensationalizing"? His analysis is very right: these companies prioritize windows over linux. That's a factual statement. Same with the collusion. This is a de-facto cartel.
ErroneousBosh 19 hours ago [-]
> This is a problem on the MSI Prestige laptop using the RTX Spark platform and Linux. Lets assume it also applies to the Asus and Dell machines using RTX spark and Linux.
Is the problem one that could be easily countered by assuming that laptops are generally better when they're not on fire?
Fr0styMatt88 21 hours ago [-]
To be fair the headline is egregiously sensationalist.
mywittyname 20 hours ago [-]
> This is a linux driver problem.
Claude, make a driver that runs at full capacity on my laptop. (make no mistakes).
user01237123674 20 hours ago [-]
To be honest, your comment is a perfect demonstration of how big corporations control public opinion by instrumentalizing people like you. Unless, of course, you work for NVIDIA.
jhowards 14 hours ago [-]
How dare they not be outraged!!
alexgoodhart 24 hours ago [-]
NVIDIA spends tens of kindly billions to *not* give their own kindly products kindly pride of kindly place
GuB-42 1 days ago [-]
No, we are just seeing a push towards Nvidia wanting to make money.
Nothing stops you from giving Nvidia half a million for their latest and greatest AI GPUs, unless you are in China I guess, but China does very well despite the export control.
It is just classic market segmentation, the thing Nvidia and just about every company that can benefit from it has done for decades. Cloud providers are ready to pay a lot of money, consumers wanting local AI have a lower budget, and gamers even lower. But a low budget is greater than zero, and there is still money to be made on these markets. But they also don't want to compete against themselves, so they are going to find ways of crippling enterprise-class workloads on consumer hardware.
It used to be that they crippled high precision arithmetic on consumer hardware, because it was a differentiating factor, but AI doesn't care about high precision, so now, it is something else.
Schlagbohrer 1 days ago [-]
I can't wait until we have real competition against NVIDIA, whether it is from a Chinese chip maker or someone else. Hopefully then these kinds of intentional crippling of hardware/firmware will be avoidable by buying an equivalent product from another firm.
FuriouslyAdrift 1 days ago [-]
AMD has been #1 for HPC and #2 in AI for a very long while. The currently shipping Helios systems go toe to toe with nVidia's next gen that aren't even shipping, yet (by nVidias own numbers).
These are multi-million dollar datacenter systems, not consumer. If you mean the consumer market, it's coming but it's going to be 2028-29 most likely (if the bubbled hasn't popped by then).
Strix Point is quite good but not the same thing as a DGX, though. Same idea, different methodology.
smallmancontrov 23 hours ago [-]
AMD's "datacenter HPC without consumer HPC" strategy was one of the biggest corporate whiffs in history. If AMD had taken consumer GPGPU seriously at any point in the decade leading up to LLM takeoff, they wouldn't have had to stand on the sidelines for years as NVDA charged 80% margin on matrix multiplication. They could have had 1/2 or even 1/1 market cap compared to NVDA. Instead, they have a tenth, and even that's largely grace in the forward looking estimations.
Fr0styMatt88 21 hours ago [-]
Haha that’s been the AMD story when it comes to graphics in general.
I miss the old days when they released enthusiast products like the R9 295X2. Though that’s just the gaming enthusiast market in general declining (not that this happened organically though, it feels like a deliberate choice).
SR2Z 19 hours ago [-]
Ha! That was certainly a card you had to be an enthusiast to put up with, with its 500W TDP :)
I have many pleasant memories of keeping myself warm just by opening Witcher 3 on my R9 390.
FuriouslyAdrift 20 hours ago [-]
It's less AMD not doing commercial GPGPU than nVidia completely staying out of the PC and server market. Remember AMD competes with both of them (nVidia and Intel) and is smaller than both.
AMDs revenue this quarter just passed Intels for the first time in history and they still are dominating in the HPC market which is fairly immune to the AI bubble (while also selling everything they can build at a high premium to AI datacenters).
AMDs margins on their Instinct datacenter GPUs is nearly as good as nVidia's. They just can't get contract volume at TSMC to make more because nVidia buys all of it.
Betelbuddy 20 hours ago [-]
You have competition now but its locked on AWS hardware, Trainium and Inferentia:
Even for things like the huawei ascend, what makes them crap for price to perf I think is still dram. They actually cost more than nvidia clusters last I checked. Just like how regular laptops and phones are now being squeezed upwards to become closer (and uncompetitive with) apple devices.
tiahura 23 hours ago [-]
China is never going to let anyone under its control _release_ AI that is capable of disrupting world order.
StopTencent 1 days ago [-]
[dead]
sidewndr46 1 days ago [-]
Given the history of prohibition in the past few centuries, it wouldn't surprise me if it is much easier to buy Nvidia products in China than in the US. The price is much higher most likely, but the actual transaction is likely easier.
smolder 1 days ago [-]
In China, you can give them a 4090 and have its ram doubled to 48GB with some soldering and custom firmware, so yeah, pretty much.
embedding-shape 1 days ago [-]
Unless the US somehow differs from the EU in some major ways, can't you just walk into a store and buy a NVIDIA GPU? Not sure how much easier it can get. And yes, it's as easy in China too, walk into store, point at GPU, fork over the money, you get your GPU. Like everywhere else, except the US I suppose?
CamperBob2 21 hours ago [-]
Yes and no. Yes, you can still buy high-end prosumer Blackwell hardware at retail in the US, but not without going through a KYC process that didn't exist a couple of months ago. The writing is on the proverbial wall.
giwook 1 days ago [-]
> but China does very well despite the export control
They're still getting plenty of Nvidia GPUs. Someone in Jensen Huang's inner circle apparently runs a number of companies in Singapore that have been found to source GPUs at scale. The catch is that she has close ties to Beijing and is suspected of working closely with their government, potentially even spying for them.
This was reported just a week or two ago and is just one example of how China is still getting their hands on resources they're supposedly unable to get.
zelphirkalt 19 hours ago [-]
There are so many ways trade goes in and out of China, and so many trading partners, I think it is next to impossible to prevent them from getting what they want, if they are willing to spend the money.
InsidiousWombat 1 days ago [-]
NVIDIA probably already figured out that local will be massive part and are preparing to do DRM for AI.
Pay up for license to unlock full capacity, pay up for license to use encrypted models from AI moguls, paid per billion tokens processed or just operations in GPU
bayindirh 1 days ago [-]
It's not their first rodeo. Their cards already need to authenticate with the driver to enable the card completely. When I last checked, Nouveau driver has received a key which only enabled the hardware, allowing enough features to see a desktop and install closed NVIDIA drivers, nothing more.
Remember, the kernel module is open source, but rest of the driver is closed source, and any kernel level code which can't be open sourced but had to be is moved to card firmware. Only the kernel module is open source. Not the libraries, not the driver, nothing else.
NVIDIA is holding to its pearls like a sloth holding to its branch, because their all business model and livelihood is dependent on that control. I still commend AMD's decision to allow full card enablement with any driver which talks the correct language.
So they can avoid the embarrassment when Copilot migrates to Qwen 4
bayindirh 1 days ago [-]
That's older than current AI/LLM/NVIDIA wave. Many application specific DSPs probably carry their model weights in an encrypted manner. If not, we would have seen the extremely performant face recognition and tracking models extracted from professional cameras already.
Note, these models generally run 60/120/240 FPS depending on the camera range, and can command the connected lenses at
~100FPS (because physics).
jval43 1 days ago [-]
AI is actually the perfect use case for homomorphic encryption.
I wonder if this space is explored fully enough at the moment.
Aside from some implementation details, it's conceptually much easier than doing it for a SQL database.
AlAjem 1 days ago [-]
I don't see the case for DRM protected local models on either side.
The consumer has access to free and open models for local use.
The supplier has better control and wider market to serve running the models in data centers.
Only way I can see encrypted models being widespread is if they are bundled as part of a software package (OS/Browser/etc) and not sold as the product.
petu 1 days ago [-]
There's good open weight models for now, labs could stop releasing them for whatever reason (assuming "good model" is a moving target).
bri3d 23 hours ago [-]
There are plenty of both transformer and non-LLM "edge" models that aren't free and open - think anything that ships in consumer electronics. Vendors usually want this IP protected.
And, the same GPUs are run in datacenter, where there's a big insider risk that vendors want to mitigate with encrypted models (if you're Anthropic and you want AWS to host your models, you don't want 5000 or even 500 or even 50 AWS employees with access to your highly valuable/desirable model weights).
selicos 17 hours ago [-]
Like locking features behind a pro card ex transcoding media (historically).
There will be ways for them to segment the market and make their money. Maybe not 100x stock in 5 years but money all the same.
sva_ 1 days ago [-]
It wouldn't work on windows either, if that were the case. It sounds more like it hasn't been properly implemented yet
Geof25 1 days ago [-]
Is DRM going to work considering how good LLMs are in reverse engineering? I would say it is a fools errand at this point.
ReptileMan 1 days ago [-]
Or buy Huawei ascend...
motbus3 1 days ago [-]
Years ago I would think I was being paranoid to think they would patch hardware retroactively to make it slower so you need to pay more to use your own hardware.
In 2026, I think this is what all car companies are doing after you pay the full car price.
I don't see a reason why they wouldn't do it with so little competition
somat 1 days ago [-]
It smells like something IBM would do. Coming from small computers where you own it, I was flabbergasted to learn that IBM would ship additional processors and ram with their mainframes but you were not allowed to use them until you paid the additional license fee.
So it is something that has been done for a long time now. Still feels like unethical rent-seeking that should not be tolerated by the customer.
See Also: car companies that charge additional fees for hardware already on the car. Additional services, I sort of understand(remote start via cell network), but stuff like heated seats, that is just wrong.
podocarp 1 days ago [-]
Databases licencing per core is the most ridiculous thing I've heard for me. In this case they have zero marginal cost and still charge you for it. Hilarious.
graemep 1 days ago [-]
> In this case they have zero marginal cost and still charge you for it.
A license has zero marginal cost in the first place. Software distribution over networks have a negligible marginal cost. An ebook as zero margin cost. We are constantly paying lots of things with zero or negligible marginal cost.
xhkkffbf 24 hours ago [-]
These deals aren't as slimy as they look. Yes, there's no marginal cost, but the goal is to spread out the costs as evenly as possible with the companies using the software. Setting the price per core just means that the companies that might do more processing shoulder more of the software cost. (And, yes, it's another way to gouge companies. But it's gouging based on use.)
graemep 23 hours ago [-]
Not necessarily slimy. The problem is that there is no clear right pricing where there is no marginal cost - makes it hard to decide whether the pricing if fair or not.
SoftTalker 1 days ago [-]
It is a case of pricing based on the value provided, not the cost of the goods. Same way we are always told to price ourselves as developers really.
motbus3 1 days ago [-]
Tesla does the same and I am not sure but someone told me Chevrolet and Audi have same "solutions" in place
esseph 1 days ago [-]
Isn't this something apple has been in legal trouble over?
> See Also: car companies that charge additional fees for hardware already on the car. Additional services, I sort of understand(remote start via cell network), but stuff like heated seats, that is just wrong.
It saves the car manufacturer on logistics cost. The per-unit cost of a seat heater is (much) smaller than the cost related to keeping two variants of seat in stock and to provide spare parts. Say the heating mat costs 100€ per seat including materials and installation, and can be sold to the customer for 400€ permanent, that means as long as 1 in 4 customers buys the seat heating option, the books are balanced, and all money you save on logistics is pure profit.
Adding subscription on top of that is financially even better, if there is one thing companies LOVE thanks to stonk market incentives it is monthly recurring revenue.
motbus3 1 days ago [-]
You are absolutely right! (Couldn't resist the ai pun)
I think more than that, it is probably a lie they tell us that you are not paying for the hardware/equipment.
somat 20 hours ago [-]
It's slimy as hell, A person buys a thing, they now own it. However there is a locked box, with no key. The manufacture sells the key to this box for an additional fee. To unlock a thing they already own.
Yeah, Nothing morally wrong there.
boltzmann64 1 days ago [-]
The reason why open source couldn't win search and social was because they need network effects which is not possible for open source maintainers.
AI is a software just like operating systems and executables where open source absolutely dominates the industry. Remember when the compiler engineers optimized the shit out of llama.cpp.
Right now the bottleneck for AI software is compute and data. Moore's law is give us compute, StackOverflow, arXiv (and it's children like bioRxiv), gutenberg and others (3lib, anna, internet.org) will give us data.
It's a waiting game. Open source always wins.
torben-friis 1 days ago [-]
>Right now the bottleneck for AI software is compute and data. Moore's law is give us compute, StackOverflow, arXiv (and it's children like bioRxiv), gutenberg and others (3lib, anna, internet.org) will give us data.
Tons of stuff happenning behind closed doors today though. No more stackoverflow for the next decade, that could create a problem if all new context only exists in chatgpt conversations only openAI has access to (or whoever).
Helmut10001 1 days ago [-]
Wait until the archives are compromised and millions of people's conversation and uploaded data are published. Such context was never before accessible in that scale, but it will become accessible soon, I am pretty sure.
oblio 23 hours ago [-]
> It's a waiting game. Open source always wins.
Based on... what? Thoughts and prayers? Feelings? Vibes?
I don't know if you looked around but open source is basically unusable for the largest platform in the world, mobile.
The next platform, whatever that is, will also be closed source (maybe VR, once it finally matures...).
LLMs are closed source, is there even 1 (ONE) LLM which is both open source and open weights and publishes its training data and methodology and is also remotely competitive with the frontier lab LLMs?
Consoles are closed source. Appliances are closed source. Embedded devices are closed source.
Even the web is basically closed source. Blobs of encrypted, obfuscated JS used to serve news pages (!).
Maybe I'm looking in the wrong places, but closed source has won.
lilbigdoot 22 hours ago [-]
What are all those mobile devices and servers running on?
oblio 9 hours ago [-]
If you think I've been on HN for more than a decade and I don't know what Android is, you're sorely mistaken. And if you think you somehow managed to reply to my comment with a "gotcha", you go digging into what Android really is made of and get back to me.
And regarding servers, who cares? The platforms users interact with are their clients and the actual server software on top of the Linux OS is closed source 99% of the time.
Don't post lazy comments.
lp92 2 hours ago [-]
This is like any other driver compatibility problem with Linux. There's tons of laptops that don't have WiFi support out of the box on Linux for example.
FartyMcFarter 1 days ago [-]
> Are we actually seeing a push towards being 100percent dependent on the cloud?
[Hanlon's Razor] This is such a contrived performance drop scenario that I find it hard to believe it was done with the purpose you're inferring.
I'd interpret this as Nvidia not caring that much about Linux desktops/laptops once again (as opposed to Windows desktops/laptops), rather than anything to do with local vs cloud.
keybrd-intrrpt 18 hours ago [-]
> I'd interpret this as Nvidia not caring that much about Linux desktops/laptops once again
Even that argument is weak - the DGX Spark is Linux, and the RTX Spark is practically the same exact chip.
Sorry to remind people of HN, but most of the consumer market still runs Windows. This was a joint project to have a consumer-level Windows-on-ARM machine, likely requiring lots of development efforts from both NVIDIA and Microsoft.
So, for these specific laptops, they put 100% of their attention for it to be on Windows for launch to capture as much market mindshare as they can. Local AI, Graphics and creative work, and some gaming.
I doubt Microsoft would have been happy about NVIDIA having Linux on their laptop right at launch.
Market Share by OS for consumer laptops & desktops:
• Windows: 76.65%
• OS X / macOS (Combined): ~15.67% (OS X at 9.86% + macOS at 5.81% due to user-agent reporting quirks)
• Linux: 5.63%
• Chrome OS: 2.04%
duskdozer 1 days ago [-]
Yes? How is this at all surprising given....everything we've seen tech companies push for for years and years? People are fully embedding themselves and their workflows in the services of big, closed companies, and are going to be surprised when they finally feel able to clamp down? There has been warning sign after warning sign, but people are only increasing their dependence.
pantalaimon 1 days ago [-]
That sounds more like the issue is a missing driver, the core does not reach the 2 GHz as it does on Windows but is stuck at 780 MHz
joshspankit 1 days ago [-]
Remember this is the same company that stopped GTX/RTX users from hardware encoding more than 2 video streams at the same time
tranceylc 1 days ago [-]
What are the implications of this?
pbasista 1 days ago [-]
I would assume that they alienated their existing customers and put off their potential customers.
4ggr0 22 hours ago [-]
not sure if the majority of customers buying gaming GPUs cares about concurrent video streams...
buggeryorkshire 21 hours ago [-]
Plenty of people like myself who have a PC as an NVR running Frigate or similar to record multiple CCTV cameras.
rkozik1989 1 days ago [-]
Yes, we are, the amount of data centers being built doesn't make sense for long-term model improvements. They are definitely planning on licensing us terminals to use instead of our own PCs at the very least as a backup plan if the frontier companies fail and have assets liquidated.
phoghed 1 days ago [-]
> They are definitely planning
Who exactly?
smolder 1 days ago [-]
Well, Sam Altman. He thinks that DRAM in personal computers is wasted and would be better used by AIs that people pay rent to access.
Beyond him, I would bet most of his bilderberg group friends agree. E.g. Peter Thiel, the owner of JD Vance.
terabyterex 1 days ago [-]
no nvidia wants you buying hardware to run local. the issue here is that the RTX spark was designed for windows.
Roark66 1 days ago [-]
Yes, and they want you to remortgage your house to pay for it.
f6v 1 days ago [-]
> Build and maintain sovereign AI while you still can.
As in, make your own inference and training hardware?
embedding-shape 1 days ago [-]
No, when people talk about building software they tend to mean building software. Just like "putting together my own training hardware" doesn't imply you're creating your own wafers from sand.
f6v 1 days ago [-]
But the original post was about a hardware vendor getting in the way. It's pointless to talk about "sovereign AI" when all you do is run GLM on Nvidia.
lcampbell 1 days ago [-]
To make an apple pie from scratch, you must first invent the universe.
Sovereign AI is less about being able to build it from first principles and more about being able to tinker and modify what already exists. I can’t train a 30B model, but I can modify the weights of one. I can’t fab my own chips, but I can putter with the software that interfaces with them.
Sovereignty is lost when encryption hides your toys and attestation tells you “you may not”.
f6v 21 hours ago [-]
Look, China has sovereign AI because they can train on Huawei GPUs.
Estonian only has sovereign AI as long as they’re allowed to buy from Nvidia. Oh wait, they’re already restricted in what they can get.
embedding-shape 56 minutes ago [-]
> Look, China has sovereign AI because they can train on Huawei GPUs.
But those servers most likely use CPUs not made in China, if I were to hazard a guess, doesn't that mean that that AI isn't "truly sovereign" in your mind then?
1 days ago [-]
hoppp 1 days ago [-]
Yes, in everything.
I stopped using my laptop for dev and switched to talking to claude code on my phone and as you can imagine I am faster at producing code.
100% cloud with a single entry point device to manage dedicated servers + agents
itsmeduncan 1 days ago [-]
This isn't as hyperbolic as people think. This is a continuation of the centralized control of data + products. Not having sovereign AI is just enabling the enshiftification of the next innovation. We let ads win, we let data mining/centralization win... can we not let 3p AI win?
hsuduebc2 1 days ago [-]
This is classical move from Nvidia. They are doing similiar bullshits on their GPUs for decades. I'm very much looking forwars for their competition to rise.
150g2 1 days ago [-]
[dead]
BalistaCRATZ 22 hours ago [-]
Unclear to me why the first attempt was to boot Omarchy, a brand new, vibecoded distro, as opposed to one of the much more reliable ones that have been around for decades
asadm 22 hours ago [-]
it's a reskin of a reliable distro no?
ctolsen 18 hours ago [-]
Bit more than a reskin now. They do enough that Omarchy distributed a configuration by default that allowed for root escalation and some UI scripts that let desktop notifications run arbitrary code. That was a month or two ago and seemed pretty vibey.
bigyabai 22 hours ago [-]
That was also my reaction to reading this. I closed the page when it became apparent the author had no idea what they were doing.
schaefer 1 days ago [-]
The article is about the RTX Spark, not the DGX Spark. The title should be changed.
sampo 1 days ago [-]
Looks like the site has fixed the title.
laughing_man 1 days ago [-]
Nvidia was already a bad choice for Linux because of closed drivers.
adrian_b 1 days ago [-]
Not really, because for more than 2 decades the closed-source NVIDIA Linux drivers were at parity of features and performance with the Windows drivers, unlike the AMD drivers before they were open-sourced, which was much more recently.
This is a new thing with the firmware that can be used only by Windows. If this NVIDIA policy continues, then NVIDIA will indeed become a bad choice for Linux, but until now, for graphics and AI applications NVIDIA has been the best choice for Linux, by providing the best performance and reliability.
For pure computational applications, especially for those requiring good FP64 throughput, there are many years since NVIDIA has become a worse choice than AMD and more recently also than Intel, but that is due to poor hardware support, not to bad software support.
lcampbell 1 days ago [-]
> closed-source NVIDIA Linux drivers were at parity of features and performance with the Windows drivers
Hell, their old closed-source NVIDIA _FreeBSD_ drivers were at parity with Windows. CUDA included (though the CUDA toolkit required Linuxulator syscall emulation to run the Linux bins).
Imagine my surprise when my new Blackwell card didn’t work, and that the newly required open source drivers were Linux-only.
laughing_man 21 hours ago [-]
Based on the numbers I looked at when I installed Ubuntu last year, while AMD cards match or even exceed Windows performance, on Nvidia a whole lot of popular games run at just over half speed compared to the Windows version, and none of them run faster.
adrian_b 20 hours ago [-]
What kind of games were those, Windows games?
If they were using DirectX through Wine, perhaps the DirectX emulation is not optimized for NVIDIA, but only for AMD.
I do not play games, but I have used professional graphics or computational programs using OpenGL/Vulkan/CUDA/OpenCL natively on Linux with the NVIDIA driver, and those have always been as fast or faster than on Windows.
The AMD and the Intel cards have always had poor drivers for OpenGL, which frequently cheated in order to get better benchmark results, by having various rendering defects in comparison with the NVIDIA cards, so when running a benchmark like SPECviewperf, the advantage of NVIDIA has always been much greater than the already higher speed, because that speed was obtained while rendering everything correctly, while with Intel or AMD many things were rendered only approximately.
I actually have more AMD and Intel GPUs than NVIDIA GPUs, but I continue to use the NVIDIA GPUs for graphics, due to a better driver, while I use Intel and AMD for general-purpose computations. However, I have not tested Intel and AMD with a Vulkan graphics application, to see whether they work better with Vulkan than with the obsolete OpenGL.
laughing_man 12 hours ago [-]
Yes, I'm mostly interested in games that were developed for Windows and running on Proton.
hypfer 1 days ago [-]
Why does the title say DGX spark when the link is the RTX Spark laptop?
Those are very different things as far as I understand, and the whole "doesn't work well with linux" really only makes sense on the RTX Spark laptop I think?
Isn't the DGX spark linux by default?
novaRom 1 days ago [-]
Good point, fixed, thanks.
19 hours ago [-]
sva_ 1 days ago [-]
It should be possible to reverse engineer this, if it works on windows?
> The laptops also support unified memory configurations up to 128GB, shared between CPU and GPU, which is unusual for a laptop and central to why they’re being marketed at AI developers in the first place.
This isn't rare on laptops, pretty default on AMD with iGPU
jcarrano 1 days ago [-]
I though the same, reverse engineer with AI. Even run the AI on the laptop and have it reverse-engineer itself.
matheusmoreira 1 days ago [-]
Won't the nvidia hardware reject open source firmware replacements? They probably only execute code signed by nvidia.
wmf 18 hours ago [-]
You don't need to replace the firmware; just send it the same commands that Windows sends.
dotancohen 1 days ago [-]
Open AI and Anthropic have already begun restricting reverse engineering with their models.
viraptor 1 days ago [-]
Fortunately that's just a matter of weeks until someone revengs and publishes the fix... even if we need to wait a while for a proper upstreamed solution.
OsrsNeedsf2P 1 days ago [-]
As someone who went down the rabbit hole of local AI, there is so much exhausting debugging already. Even if this is "just" another step you have to find, it's toxic.
Neywiny 1 days ago [-]
Yeah I'm giving all my RAM to Sam Altman tomorrow because of that
AdamConwayIE 1 days ago [-]
I wonder if you could port the GB10's drivers over? They're extremely similar and the GPU itself identifies all the same specs.[1]
Also very curious if the NPU is visible in Linux, too.
It's not clear to me if this is just an egregious bug or intended malicious behavior? I like to generally give people the benefit of the doubt.
nonethewiser 1 days ago [-]
How is everyone wildly misunderstanding this article?
It's lack of linux driver support for firmware designed for windows only laptops.
What is going on here? Commenters here are clearly extremely paranoid about NVIDIA centralizing inference. This is not reflected in the facts nor narrative of the article - it's completely fabricated out of left field her in the comments.
This is a problem on the MSI Prestige laptop using the RTX Spark platform with Linux. Lets assume it also applies to the Asus and Dell machines using RTX spark and Linux.
RTX Spark is a new platform and all of these laptops ship with Windows. You cant get Linux on any of these machines from the factory. NVIDIA may even ship drivers for linux later as they did with Dynamic Boost.
21 hours ago [-]
rovr138 1 days ago [-]
the slowdown might be a bug, the encryption feels malicious in the 'controlling' way after buying it.
TkTech 20 hours ago [-]
Most nvidia consumer/desktop GPUs are crippled intentionally in various subtle ways - tensor cores for example are capped at half speed on all GeForce cards, which is why https://github.com/Zagitta/gpu-unlock exists. The intent is to force you to spend significantly more for the exact same silicon under commercial packaging if you want to do inference tasks.
smlacy 22 hours ago [-]
NVidia has had encrypted/signed driver binaries on linux for ... forever?
I understand that their "secret sauce" might be exposed with open drivers, so I somewhat respect their right to intellectual property in this respect.
heeen2 22 hours ago [-]
they have had encrypted firmware for a long time already, see the missing clock management in nouveau.
juujian 1 days ago [-]
This is unfortunately the normal release cycle for Linux, especially with Nvidia. If you buy the latest and greatest, some features just aren't available. Eventually, manufacturers work with Ubuntu to unlock some of those features and other distributions then adopt those drivers. Takes a few months though. If you're lucky, the drivers become part of the Kernel, and that process takes even more time. Some features never land at all (looking at you, AMD XDNA gen 1).
petercooper 1 days ago [-]
Assume I'm naive on this matter, but why wouldn't Nvidia allow its own DGX OS to be a first party OS on these devices? Some sort of MS subsidy? Ecosystem-wise, wouldn't it make sense for Nvidia to keep pushing its own OS on all the devices containing its hardware?
lolive 14 hours ago [-]
Isn’t this RTX product targeted at the corporate market, proposing privacy and local AI for laptops, targeting an audience that is wealthy, massively uses Windows [and may be WSL], and is in dire need of that feature, so willing to pay.
impossiblefork 1 days ago [-]
Unrelated to this slowdown, I could imagine this sort of thing being the right application for things like Taalas, although I'm not quite sure whether it can actually offer what I imagine that it can (i.e. ability to make a cheap, physically small card that can run a large model).
A laptop can't run the latest big models, because memory is too expensive. This thing probably can't run models larger than Mistral Small 4. But it'd be a completely different thing to have a hardcoded Mistral Large 4, than a choice of tiny models.
wmf 22 hours ago [-]
Taalas won't be cheap. Each chip will be at least $1,000 and large models require multiple chips (maybe hundreds).
RealityVoid 1 days ago [-]
You probably won't be able to run a sota model on a single Talaas card. You probably need a set. It should theoretically be cheaper, faster, more power efficient than Nvidia at running that model. But then you are stuck with a chip that can run only that model. And I understand that context size will also be a problem.
impossiblefork 1 days ago [-]
Yeah, that was why I felt that, while Taalas might not be the right thing in general, it could be really appealing in a laptop, that would otherwise not be able to run any big model at all.
I think I misunderstood what the technology could do a little bit, because I imagined these fixed matrix multiplies encoded into hardware as small-ish, as using routing space etc., not just fast.
I think what I really imagined was useful would be more easily achieved by having a big ROM with model weights connected to an ordinary GPU.
nubinetwork 1 days ago [-]
> the GPU gets locked to roughly 780 MHz on Linux instead of the 2 GHz-plus it hits on Windows
This has been an issue on Linux since 2014, and the gtx980.
jasomill 1 days ago [-]
Unplugging a PC containing a Maxwell+ card doesn't properly set its clock speed.
Even if this is an intentional nerf by Nvidia, which I doubt, then then nerf is buggy.
nubinetwork 21 hours ago [-]
The GPUs start up with the lowest speeds, and only the binary driver has the bits needed to "reclock" the card. That's partially why the oss driver sucks these days.
throw93030e79 1 days ago [-]
2 months old laptop has power management issues on Linux. What a suprise!
Still miles ahead compared to Apple, where linux does not even boot on newer macs!
PcChip 1 days ago [-]
Agreed, i bet they will fix this quickly to help sell more units
khgihpipihkjdsg 1 days ago [-]
I am starting to wonder whether CEOs everywhere ( but especially in US ) are simply unable to read the room. I shrugged off nvidia when they did it to gaming and mining ( it annoyed me, but ultimately there were options to deal with it ), but here it is touching a much wider populace with repercussion well beyond crypto and gaming.
barrkel 1 days ago [-]
I remain puzzled why anyone would want to do inference on a laptop if they have a home internet connection and a phone.
If you really want to do DIY inference, plug it in at home and set up tailscale so you can access it on the go.
No huge battery drain or toasty hot laptop required. Lower cost laptop. Lighter laptop. Longer battery life laptop. Win win win win.
solyin 1 days ago [-]
Not everyone grew up in a situation where home was a stable dependable place. Not everyone has a home to go to. If you have it with you, they can't mess with it or take it away from you when you're not there. If you grew up in a broken home, you would understand.
Gigachad 1 days ago [-]
And those people are buying $10k laptops to run local AI?
jeroenhd 1 days ago [-]
Those people won't be buying the Spark laptop, but asking "why anyone would want to do inference on a laptop" is broader than what the article covers.
c7b 1 days ago [-]
Local AI is an expensive hobby. You're paying big money for small performance (so mainly for the feeling of "owning it", which is a great motivator). It can't match cloud services, not absolutely nor bang-for-buck, especially if you look a bit for cheaper and freebie options. As someone who loves and promotes local AI, in the situation you described my plea would be to use that money to improve your situation more directly (move out, get support,...). The cost of a laptop that can run somewhat decent models should buy you at least several months of therapy (@1 session/week).
throw1234567891 1 days ago [-]
You can't put a price of sending all your IP to a dubious third-party. Sorry. You argument apples vs oranges.
c7b 1 days ago [-]
If you want local AI, which I absolutely encourage you to, I agree with the OP. Get a (mini)-PC and set up a VPN. It's going to save you money compared to a laptop, short- (usually cheaper) and long-term (usually better upgradeability). You'll have more money for therapy, which I would also encourage you to consider if you come from a broken home, as the GP mentioned.
throw1234567891 1 days ago [-]
Sure, they pay you well at Anthropic or another OpenAI to go and ridicule people? Not everyone builds some irrelevant SaaS where everything is for sale. Some of us work with other people's data and treat it seriously.
Maybe you need a therapy. FOMO can be treated, too. Or you're just upset about Trump falling into irrelevance.
c7b 1 days ago [-]
Maybe you're responding to the wrong thread. The discussion was about laptop vs PC for local AI, which should be pretty equivalent from the perspective of Anthropic, in the context of a broken home. Read the previous chain and calm down.
throw1234567891 19 hours ago [-]
Thanks, sometimes it's good to hear: stop. I'll think about the therapy.
Gigachad 19 hours ago [-]
The situation we are talking about here is someone with no fixed home or insecure housing. That kind of person likely isn’t even paying for cloud AI, let alone buying massively overpriced hardware to run it locally.
The market for someone who wants local AI in a laptop for $10k has to be tiny.
1 days ago [-]
Danox 1 days ago [-]
The spark laptop in its present form is not very good, in fact, it’s disappointing.
barrkel 1 days ago [-]
If the security of your home possessions is an issue, it should be addressed first before you spend thousands on a very stealable laptop that you need to carry everywhere.
dirasieb 1 days ago [-]
homeless RTX Spark owners, that's a new one
solyin 21 hours ago [-]
You grow up, that mentality still stays with you. I know multi-millionaires that do the most broke ass poor people nonsense because they grew up poor.
selimthegrim 22 hours ago [-]
How many grad students have you met?
alerighi 1 days ago [-]
Because maybe you don't want to send your private data to a company that could use it for whatever it wants?
barrkel 1 days ago [-]
Can you explain what data is going where?
entrope 1 days ago [-]
Your prompts and whatever data they collect from your dev environment go to the inference provider, and typically the customer has neither control nor assurance over what the inference provider (re-)uses that data for.
Orygin 1 days ago [-]
Your prompts and data are going through your own VPN to your own inference hardware. You did not read the post apparently
anonym29 1 days ago [-]
GP isn't proposing using closed proprietary models over an API, they're proposing having local hardware serving open weight models at your house/apartment being remotely accessible through an overlay network like Tailscale. The idea being that local inference is very hungry for power and thermal budget, so you run the inference somewhere that power and thermal headroom is abundant (your house/apartment, on a tower, or dedicated unified memory boxes like DGX Spark, Strix Halo, Mac Studio, etc) and then remotely utilize that inference from a regular laptop. You still get all of the benefits of local, private AI inference, without having to expend a laptop's limited power and thermal budget on doing the inference on the laptop itself.
spaceywilly 1 days ago [-]
On the contrary I think in a few years most inference will be done locally. The smaller models have been slowly catching up to frontier for many tasks, and chips are being developed to run large models on phones and laptops (according to Qualcomm CEO they want to run 10T weights on a mobile chip).
To me the simple answer to why is control. If I run a model locally I get to control the model params, the system prompt, all of it. And of course there is no quota. But probably most importantly I am not tied to paying some company a monthly subscription forever.
KeplerBoy 1 days ago [-]
The RTX Spark is also sold in a mac mini form factor. That's pretty well suited for home inference.
barrkel 1 days ago [-]
Exactly. Get yourself a DGX spark and plug it in at home.
I meanwhile, run my ridiculously OP workstation class hardware as a server - 9995wx, 768GiB, rtx pro 6000 w/ 96GiB vram - at home, and my Asus ZenBook is light, cool, and long-lived.
Runs GLM 5.3 flash at 35 tok/sec while consuming only 15 GiB VRAM, and I can run a lot more stuff on there alongside.
jacquesm 1 days ago [-]
That's slower than I would have expected for the RTX 6000, do you have figures for say Qwen 3.8 27B? That model fits completely on the card, that should push it a lot further.
barrkel 2 hours ago [-]
I'm pretty certain I can get it to go faster by caching experts and whatnot (with experimental llama.cpp branches), but that would consume a lot more VRAM and it's fast enough as it is for playing around with, and I can leave it always loaded. It's like 50% CPU inference, with a few layers + router + kv cache on the GPU.
dlahoda 1 days ago [-]
private memory.
unique local data and know how not shared with cloud.
barrkel 1 days ago [-]
What part of what I said implied data was going to the cloud?
pixelpoet 1 days ago [-]
I have been saying this for years, and the answer hits a nerve that gets you downvoted into oblivion: people prefer the aesthetics of laptops, simple as that. They will absolutely go to the grave arguing that they need the mobility etc, then just leave it plugged in. They simply prefer how laptops look, and they absolutely hate it when someone dares say it.
monegator 1 days ago [-]
I have a laptop for work. 95% of the time i'd prefer a tower or at least a mini PC with proper cooling, but that remaining 5% i have to be around with the laptop is enough. I don't like the idea of having to remember to sync the projects every time (especially while i'm in the middle of testing something new so it's not like i can just pull the latest commit or sync the latest good version of something)
dumberquestions 1 days ago [-]
If you do any travel whatsoever a laptop is usually not optional.
DaSHacka 20 hours ago [-]
A laptop that is so beefy it can even run GLM 5.3 locally (while sounding like a jet engine, being as bulky as an old-school travelling briefcase, and having a 1.5 hour battery life) though?
GGP's point is just: keep the bulk of your compute at home, where airflow and energy is cheaper (in an opportunity-cost sense). Keep the laptop portable, cool, and light, with long battery life by offloading your own computing tasks to your own hardware.
barrkel 1 days ago [-]
To be absolutely crystal clear:
I am not saying don't use a laptop.
I am not saying you have to use cloud inference.
I am saying using a laptop for inference is dumb. If you want private inference:
- use a dgx spark or similar plugged in at home
- set up tailscale so you can connect from anywhere
- use your phone as a hotspot on the go
- work on your laptop, using tailscale to connect to your home inference server if you're on the go
This gets you everything except for the few hours you might be on an airplane.
pixelpoet 1 days ago [-]
100% agreed, and people saying they need a laptop are missing the point; I wasn't saying don't use laptops, either.
The simple point I'm always trying to make is that area vs volume is a fundamental physical difference regarding cooling, which is critical for high performance applications (that's why they're considering an RTX Spark in the first place, right?).
There's so much more I could say about my Strix Halo machine, which fits in the palm of my hand and I have a wonderful portable screen and keyboard+touchpad (and only cost me 2k with 128 GB fast memory, idles at 4w, amazing 32 thread Zen 5 CPU with AVX-512, ...), but it's either preaching to the choir or pissing against the wind.
1 days ago [-]
bitbasher 1 days ago [-]
We can reverse engineer anything... why not start with Nvidia's drivers and so on?
wewewedxfgdf 1 days ago [-]
Remember when Apple was slowing down iPhones, the outrage!
But with Nvidia people just expect they nerf all the features.
wmf 18 hours ago [-]
The whole point of this article and thread is that people expect it to work and are mad that it doesn't work.
nibbula 1 days ago [-]
You will never stop me. As I and everyone like me, before and after me, have shown, time and time again. And by me, I mean nature. And by nature I mean math and logic unchangeable in any possible universe. You cannot give me a thing without me having it. If a thing exists, I can copy it, know it, and bring it's equivalent into existence. Nothing can be created that is not indelibly written into the the intricate cause and effect chain of history. A perfectly encrypted thing is indistinguishable from the noise of the cosmos, but if I want I thing, it shall soon emerge, free from any control from that noise, which is the weight of all that has happened, just as surely as I can count to a number.
If you want to prosper with the current of nature, provide a thing easily for mutual benefit. The more you try to block a thing, the sooner your demise into uselessness.
Is this really news? I thought the community has long ago conceded defeat to Nvidia on this issue since the Nouveau driver as well as the new Nova drivers will both utilize Nvidia's firmware blob.
999900000999 1 days ago [-]
>Oddly, the direction of the problem flips depending on power source. Unplugged, Linux hit around 61 to 67 tokens per second on LM Studio, comparable to Windows. Plugged in, that same test dropped to about 15 tokens per second.
Absolutely hilarious.
This is just incompetence. No reason Nvidia with its TRILLIONS of market cap couldn't of tested and fixed this on Linux before shipping.
Or a day one patch.
I have no doubt that'll it'll work right eventually. Long after RTX Spark laptops flop and can be purchased for half off.
Last night I pulled out my 5070TI laptop and installed CachyOS. I was running various models within about 20 minutes.
That 12GB of VRAM is a hard limitation, Unsloth can't split jobs between vram and system memory.
throw1234567891 1 days ago [-]
> couldn't of tested
Could have tested.
anonym29 1 days ago [-]
Thanks, I had absolutely no idea what GP was saying, it was completely incomprehensible until you published this Rosetta Stone here.
throw1234567891 1 days ago [-]
Enjoy.
crest 1 days ago [-]
What makes you think this was an oversight?
999900000999 1 days ago [-]
Because they also could of just locked the bootloader fully if they didn't want Linux to work.
Even from the article I'm guessing you could write a script to automatically extract the encryption keys while on Windows and the load those in Linux.
This is only mildly inconvenient for anyone who'd use this in a data center cluster.
quaffcode 1 days ago [-]
We have antitrust laws on the books. They have barely been enforced since Reagan. But that good change if there was political will. How is Nvidia not a monopoly?
wmf 18 hours ago [-]
Nvidia is certainly not a monopoly in laptops.
jacquesm 1 days ago [-]
Maybe they should try installing the same distro that is used on the DGX spark to see if that works.
Zak 1 days ago [-]
> the firmware’s power table is encrypted and only the Windows driver can decrypt it
Why the fuck is the power table encrypted?
a-dub 1 days ago [-]
hm. nvidia is garbage on linux again. just as the prophecy foretold. we are safe from the singularity, for now.
throwawayffffas 1 days ago [-]
I do wonder how much it would cost to "clean room" engineer the firmware using AI.
porphyra 18 hours ago [-]
honestly the fact that right at release it works with
> laptop ran a full desktop at 4K and 120Hz, with working audio, touchpad, keyboard, and CUDA acceleration
is amazing. It seems that this person's MSI version had some quirks with firmware but Omarchy already supports the Asus and Dell versions of the RTX Spark out of the box. I'm surprised that everyone is focusing on the negative when, prior to RTX Spark's release, it wasn't clear if Linux would work on the RTX Spark at all in the first place.
This is way better support than Qualcomm's Snapdragon X1 and X2 laptops, which are still struggling to get Linux support years after release.
Presumably, once this one power management driver issue gets sorted out (surely an accidental bug rather than what some of y'all cynics are suggesting), it will be great. Now if only it could be cheaper...
dagmx 21 hours ago [-]
As an aside, I made the mistake of buying the Surface Laptop Ultra and it’s such a dud.
NVIDIA and Microsoft tried so hard to make a Mac clone with this and it’s just missing the mark everywhere.
It doesn’t do the dynamic memory split that a Mac does so out of the box it reserves 32GB (half) for the GPU and windows limits CPU processes to that. Getting more memory requires a reboot.
And because I want Unix tools (without Cygwin), WSL2 splits that even further in half.
Performance is way behind a similarly specced Mac. The GPU is thermally throttled near a 5060 laptop. The CPU out of the box came in a lower power state but windows power management makes it stuttery unless you change that, at which point it’s cooling is a hurricane compared to the Mac which outperforms it while being quieter.
The new MagSafe like connector is stupidly designed too. You can’t generically use it in any USB-C port, it’s super chonky and disconnects only on the horizontal axis. Worse, because it’s a right angle connector it can only be used in one orientation without blocking the other ports.
Windows input handling sucks too. There’s a lag before the trackpad responds based on moved distance. Some terrible take on palm rejection I’m sure. But even their display is mediocre because it gets very bright but they messed up local dimming. So if you select a bunch of text in the terminal it blinds you.
Idk I got this specifically to do windows dev with CUDA on arm. So I knew it was niche but anyone else buying this is making a mistake. It’s an unfinished product trying to be a competitor that nobody in Microsoft even understands.
This Linux support issue is just the icing because everyone keeps asking me why I don’t just put Linux on it and workaround my issues, but you can’t.
I’ve spoken to friends at NVIDIA who say it’s rushed out. So maybe software will get better but I urge folks to not buy this unless you’re into the same niche needs as me.
skavi 21 hours ago [-]
> It doesn’t do the dynamic memory split that a Mac does
Odd because Tegra/Jetson platforms have been doing this for years.
dagmx 21 hours ago [-]
I believe the DGX Spark under Linux does it too. Which is an otherwise identical chip except for the networking.
I suspect it’s a limitation of them focusing on Windows driver and windows being far behind the curve.
blks 1 days ago [-]
Nvidia being the shittiest hardware vendor for GNU/Linux, as always.
aa-jv 1 days ago [-]
As a DGX Spark user, this article title gave me conniptions.
But, on the other hand, having my DGX Spark stack at hand, I know the solution:
/btw reverse engineer the Nvidia Platform Controllers and Framework component and build a Linux version
"AI views proprietary components as damage and routes around them."
K0balt 1 days ago [-]
The attack on reverse engineering of games is a precursor to universal computing lockdown. When you need AI to keep up with programming, the entity controlling the AI controls what software is possible to create. It’s a bid for centralised control and thought censorship at a scale and depth much more ambitious than anything we have ever seen before.
Sovereign AI has never been more important.
aa-jv 1 days ago [-]
The ecosystem and community around the DGX Spark platform is semi-sovereign AI, or at least its as close to it as we can find so far.
But yeah, we do need to make sure the REA cat stays out of the bag.
I, for one, welcome our reverse-engineer-anything overlords.
FartyMcFarter 1 days ago [-]
> conniptions
TIL. Nice word.
u1hcw9nx 1 days ago [-]
Do you know that 1 + 1 = 10 in binary?
aa-jv 1 days ago [-]
I know that, because I read the article, but if all I'd read was the title, I'd not know that.
EDIT: I don't know why you changed your comment, but yes, it was obvious from reading the article that it was not actually about the DGX Spark.
stevefan1999 1 days ago [-]
This reason alone is why I rather stay on Strix Halo
rich_sasha 1 days ago [-]
This one is allegedly/hopefully a bug but just waiting for it to become a revenue stream / paywall. Uh oh, you want to run locally your dangerous illegal foreign open models, not licensed by Nvidia? Can’t work with you paying for a special security feature unlock on your GPU.
Or maybe just not at all - shell out for OAI tokens and send us all your data - your privacy is very valuable to us and we will make sure to be the only ones to monetize it.
You could say, Nvidia is selling pickaxes and has no business gatekeeping what rocks you hit with it, but I guess it depends on their relationships with major mining companies, and ultimately what pays more.
ilaksh 1 days ago [-]
So weird that everyone gives Microsoft a complete pass. They probably paid Nvidia to give them a head start so people would be forced into Windows and WSL for awhile.
Zigurd 1 days ago [-]
Elsewhere on this thread someone mentioned a Microsoft patent that appears to be related to this topic. I think you've found the thread that needs to be pulled here. With their background in the game industry I wouldn't be surprised if Nvidia is comfortable with this kind of fuckery.
1 days ago [-]
demo4567 1 days ago [-]
Next GPU is going to be AMD
vermaden 20 hours ago [-]
Nopevidia strikes again ... just use open solutions from AMD and/or Intel when it comes to GPUs.
Vote with Your wallet.
Its that simple.
Every.
Single.
Time.
Hizonner 1 days ago [-]
Maybe stop buying NVIDIA's crap now?
nekusar 1 days ago [-]
And anybody who watches proprietary APIs like CUDA saw this coming for, like, 15 years.
They've always played these sorts of games. Now, people are seeing the substandard solutions for what it is, and are angry.
yieldcrv 1 days ago [-]
Decompile time
alerighi 1 days ago [-]
As always, nothing has changed, Nvidia is always shit in supporting Linux.
m00dy 1 days ago [-]
just reverse engineer the firmware...
villgax 1 days ago [-]
much ado about nothing, it'll probably come out in some time. Just microsoft trying to capture market/midnshare initially
TobiasJacob 1 days ago [-]
Wtf is Nvidia doing?! I was considering buying one, but now?
They really got it right on multiple levels - for those who don't know, roller coaster tycoon was infamously written in 99% assembly and meticulously optimized by the developer to run effectively in 16 megabytes of ram.
dzonga 1 days ago [-]
ai labs need better stewards besides the carton / comic level villainy from sam altman and dario
totetsu 1 days ago [-]
I'm starting to feel empathy with those people who think bus lanes are coming form their freedom as part of some conspiracy to take away their cars.. I would like private enjoyment of my own RAM to use as I wish please.
shevy-java 21 hours ago [-]
Or alternatively it is time to help support companies that support
Linux. As Nvidia keeps on being a problem, let's look for alternatives.
We can see with this evil cartel currently driving up the RAM prices
how greedy they all are. Our money should go into fairness. The US
court system has given up on that. Even on capitalism. Capitalism used
to claim that a free market with competition leads to lower prices.
Well - we have several de-facto cartels here. And nobody is breaking
any of these up.
saidnooneever 19 hours ago [-]
cmon Linus, give em the finger again !
classified 1 days ago [-]
It was only a question of time until enshittification would arrive on GPUs. This will continue and expand, since there are no consequences.
This is a lack of linux driver support for firmware designed for windows laptops.
This is a problem on the MSI Prestige laptop using the RTX Spark platform. Lets assume it also applies to the Asus and Dell machines using RTX spark.
RTX Spark is a new platform and all of these laptops ship with Windows. You cant get Linux on any of these machines from the factory. There is also precedent for NVIDIA shipping linux support well after Windows (2 years in the case of Dynanmic Boost, a similar power sharing feature).
It is not obvious to me why they would want to do that.
They basically do. ChromeOS runs a Linux kernel, so the userland has everything it needs to run CUDA. FreeBSD is a first-class target for Nvidia drivers, and macOS used to be supported too but stopped signing Nvidia code. Nevertheless, AArch64 is still supported for CUDA et. al..
It is up to Nvidia to support whatever platforms they want users to be on. In their case, it means supporting many OSes.
My point is that's not Nvidia's responsibility. The article itself admits the laptop "ships with Windows on ARM", so support for another OS shouldn't be expected, even if NVIDIA supports other platforms and other product lines differently.
There are OEMs that refuse this responsibility. If that's the vindication you wanted, there you go. This specific issue is not a catastrophic oversight though.
Personally, I don't care. If you don't support open drivers, you have to provide first class drivers on Linux on day 1, because NVIDIA's majority of business is running on Linux already.
On top of that, they claim to have the best Linux support, and claim to have best drivers in their class.
So, it's better to act like that and provide support. If any other company has done that, they have been incinerated with the flame of internet's hate thrice already.
> There is also precedent for NVIDIA shipping linux support well after Windows
There is also a precedent for NVIDIA nerfing drivers on Linux and pulling anti-competitive shenanigans against competition, so it's only reasonable that people look at the "missing" driver support with suspicion.
How does your conspiracy theory hold up to that?
GB10 and N1X are both Blackwell cousins. So they've had consumer-level Linux machines for a while now.
This was simply a partnership with Microsoft, so of course development of drivers and firmware focused heavily on Windows. Especially since I'm sure Windows required a LOT of updating to run nicely on ARM as well as the N1X unified memory model.
You may as well be right, I am speculating and this theory might be baseless (I don't think it is.) But GP's comment is wrong to ask people to simply read the article at face value. That is bad reading.
Then you misunderstand how hardware, software, firmware, drivers, and operating systems have worked for the last few decades.
Good luck with that. A PC is more or less defined by specifications laid out in Microsoft's Windows Hardware Compatibility program and tested by their Windows Hardware Lab Kit. Before that, it was defined by IBM.
I've heard that ARM has had a standard for ARM-based servers, their Server Base System Architecture, but I've never seen any compliant hardware in the wild.
I don’t think anyone “allowed” them - quite the opposite. They’ve made generational bets that ended up being right, bets that IIRC the industry did not agree with for a long time. Them pursuing CUDA resulted in a draw down of something crazy like 75% of their valuation. They bet the farm multiple times to get to the point they are - and it’s only “obvious” in hindsight.
If you want folks to make generational bets as a matter of course (which I think most folks do), centralizing the downside of those bets and socializing the upside of those bets is not a great strategy. Almost as bad as socializing the downside.
Competition will eventually close that gap. We’re only a few years into this thing.
Have you ever heard of U.S. Steel, IBM, Kodak, and Western Union?
https://developer.nvidia.com/blog/nvidia-transitions-fully-t...
Point was about the kernel drivers being open-source enabling hackability - they don't, the card hardware is just as closed as before.
When did "we" have a say in this? I don't get these types of comments from the "open-source as a community."
Nvidia has been doing this for decades, and there are two possible arguments:
1. They are successful despite it.
2. They are successful because of it.
There is no mythical collective "we" representing the open-source community in Nvidia's supply chain of success. It's Nvidia and its consumers.
Nvidia won because their architecture is better. If you do a significant amount of work that can be CUDA-accelerated, you're kinda wasting your time not owning a CUDA-capable card. It's like being a video editor that software-encodes their projects because they don't want to get an accelerator card. It's your choice, but software-acceleration is a nightmare compared to how quick and efficient hardware-acceleration is. Competitors will all use the accelerator card and ship faster than you can.
Apple, AMD and Intel have all earned their current position by abandoning Khronos and refusing to build their own GPGPU architecture. If the entire industry refuses to invest in complex technology just because it's expensive, then they don't deserve to make money.
I was genuinely excited about this device until I heard the specs. Still under 300GB/s memory bandwidth when a gpu like rtx3090 from 6 years ago had 900+
This is really bread crumbing local inference users.
We should be measuring our memory bandwidth in TB by now. The only company that seems to be serious about this is Apple. (the m5 ultra studio etc).
Or RTX Spark and run their workloads under WSL, which as the article notes, has minimal overhead relative to running Linux directly?
This sounds more like a bug than a feature, especially given that Linux on battery runs at full clock speed and downclocks when plugged in, and Windows does the opposite, suggesting they could at least mitigate the problem with no changes to their software stack by inverting a conditional in the firmware and adding a Linux kernel parameter to control memory allocation as apparently already exists for Windows.
The only value proposition of RTX Spark at all is the laptop form factor
In fact I'm pretty sure it's happening. Why? Because of constantly growing prices of those things.
> NVIDIA has now confirmed an official starting price of $2,399 for RTX Spark laptops
https://shattered.io/nvidia-rtx-spark-mediatek-oems-price-20...
https://www.servethehome.com/nvidia-dgx-spark-64gb-launched-...
If we look at what you can actually buy the 64GB RTX Spark is $4,500 and the 128GB one is $7,000: https://eshop.asus.com/us/31150-us-cnf-auto.html & https://www.dell.com/en-us/shop/laptop-computers/spd/xps16dx...
Hell there doesn't seem to be an RTX Spark that's sub $3,300 when considering the 32GB models even. $2,399 is pure fiction.
And if there's a market for a DGX Spark configuration with a higher compute-to-RAM ratio, why isn't Nvidia offering it?
My best guess is that the entry-level Sparks are mostly intended for people who would otherwise be in the market for a traditional high-end laptop, who might be enticed to spend a few hundred more for at least the promise of local AI, while the higher-end models are intended for more serious AI development.
With that said, it's not hard to imagine there isn't nearly as large a built-in market for high-end SFF Linux PCs as there is for high-end Windows laptops.
Nvidia does something to prevent abuse so that regular people who want the laptops can buy one. Nvidia gets knocked on.
This is all but certainly a bug in their linux driver and nothing else. It sure as hell isn't some "anti abuse" to enable "regular people" to buy one.
Man it really is exhausting to read a title that sparks my interest, just to be greeted with the most basic Claude-isms in the subheading.
I see what you did there. ;)
I just heard a quick J. Talarico speech at some country concert. In the 10 or so sentences he spoke, he had 3 or 4 instances of doing this "Claudism" enumeration. Are you saying Texas just elected Claude for Senate?
Maybe we should stop with such superficial judgements. It's the modern equivalent of not liking the speaker's tie. (Doesn't mean the speaker's speech is good, but it doesn't mean it's bad either.)
Edit re "equivalent to not liking the speaker's tie": This is not equivalent at all, this is a written blog post and I'm put off by the way it is written, not by say the URL or their chosen font. It's more akin to a speaker dropping an inappropriate joke at the start of their talk and leaving because of that
Sovereign AI has never been more important.
Controlling AI is so much more powerful and impactful than controlling search or social algorithms.
Centralized AI a 10x on social control over everything we have seen before, and there is no way that a free and open society survives contact with centralised, high capability AI if the system is manipulated towards guiding outcomes. It’s essentially mind control, as people will offload an even greater share of their cognition to AI than they do with search and social.
Build and maintain sovereign AI while you still can.
This is a linux driver problem. On laptops designed for windows.
Read the article and stop sensationalizing.
This is a problem on the MSI Prestige laptop using the RTX Spark platform and Linux. Lets assume it also applies to the Asus and Dell machines using RTX spark and Linux.
RTX Spark is a new platform and all of these laptops ship with Windows. You cant get Linux on any of these machines from the factory. NVIDIA may even ship drivers for linux later as they did with Dynamic Boost.
Kindly remove this from your list of "Evidence that NVIDIA is Colluding to Centralize Inference" and everyone else take note of this confirmation bias in action.
edit: even more important for mobile computers
That them trying to protect IP around something very profitable isn't crazy. That not being fully supported in Linux, when the product was literally designed around Windows, also isn't crazy or really evidence of anything, including evil (see the whole fucking history of Linux compatibility of made-for-Windows hardware).
Maybe let them cook.
It obviously can't be because of ulterior motives!
nvidia has always been crap like this. Look at amd and intel, they let u write proper drivers (even if its real difficult to do so). No encrypted blobs or voodoo needed.
the nvidia drivers u need to use a card on Linux are unintelligible binary blobs. thats sad as it is. this is worse. they might give some shitty blobs at some point, after they made enough ppl give up on it and switch to give them and their partners their money, souls and firstborn
No. It's 'the NVidia way or the highway' problem. NVidia is so full of themselves they refuse to work under anyone's rules (i.e. the Linux kernel where they're constantly trying to work around them).
They're egomaniacs who are so full of themselves, riding high on the AI bubble, they think they're above everyone else in the ecosystem, while utilising all kinds of other people's work for their own benefit.
Look at the way they develop their 'open' Linux drivers where it's just a shim that loads a huge binary blob off the GPU rather than a proper driver like amdgpu has.
Meanwhile they wouldn't be a $5T company without Linux and the wider FLOSS ecosystem.
It's kind of like when Secure Boot debuted and the plan was to allow only Windows with it enabled, then no doubt someone would claim the fact that only Windows can boot with it enabled as a 'Linux driver problem'.
Is the problem one that could be easily countered by assuming that laptops are generally better when they're not on fire?
Claude, make a driver that runs at full capacity on my laptop. (make no mistakes).
Nothing stops you from giving Nvidia half a million for their latest and greatest AI GPUs, unless you are in China I guess, but China does very well despite the export control.
It is just classic market segmentation, the thing Nvidia and just about every company that can benefit from it has done for decades. Cloud providers are ready to pay a lot of money, consumers wanting local AI have a lower budget, and gamers even lower. But a low budget is greater than zero, and there is still money to be made on these markets. But they also don't want to compete against themselves, so they are going to find ways of crippling enterprise-class workloads on consumer hardware.
It used to be that they crippled high precision arithmetic on consumer hardware, because it was a differentiating factor, but AI doesn't care about high precision, so now, it is something else.
These are multi-million dollar datacenter systems, not consumer. If you mean the consumer market, it's coming but it's going to be 2028-29 most likely (if the bubbled hasn't popped by then).
Strix Point is quite good but not the same thing as a DGX, though. Same idea, different methodology.
I miss the old days when they released enthusiast products like the R9 295X2. Though that’s just the gaming enthusiast market in general declining (not that this happened organically though, it feels like a deliberate choice).
I have many pleasant memories of keeping myself warm just by opening Witcher 3 on my R9 390.
AMDs revenue this quarter just passed Intels for the first time in history and they still are dominating in the HPC market which is fairly immune to the AI bubble (while also selling everything they can build at a high premium to AI datacenters).
AMDs margins on their Instinct datacenter GPUs is nearly as good as nVidia's. They just can't get contract volume at TSMC to make more because nVidia buys all of it.
https://youtu.be/Bteba8KLeGc
https://d1.awsstatic.com/onedam/marketing-channels/website/a...
https://youtu.be/5SgZ3zRJFAY
They're still getting plenty of Nvidia GPUs. Someone in Jensen Huang's inner circle apparently runs a number of companies in Singapore that have been found to source GPUs at scale. The catch is that she has close ties to Beijing and is suspected of working closely with their government, potentially even spying for them.
This was reported just a week or two ago and is just one example of how China is still getting their hands on resources they're supposedly unable to get.
Pay up for license to unlock full capacity, pay up for license to use encrypted models from AI moguls, paid per billion tokens processed or just operations in GPU
Remember, the kernel module is open source, but rest of the driver is closed source, and any kernel level code which can't be open sourced but had to be is moved to card firmware. Only the kernel module is open source. Not the libraries, not the driver, nothing else.
NVIDIA is holding to its pearls like a sloth holding to its branch, because their all business model and livelihood is dependent on that control. I still commend AMD's decision to allow full card enablement with any driver which talks the correct language.
Note, these models generally run 60/120/240 FPS depending on the camera range, and can command the connected lenses at ~100FPS (because physics).
I wonder if this space is explored fully enough at the moment.
Aside from some implementation details, it's conceptually much easier than doing it for a SQL database.
The consumer has access to free and open models for local use.
The supplier has better control and wider market to serve running the models in data centers.
Only way I can see encrypted models being widespread is if they are bundled as part of a software package (OS/Browser/etc) and not sold as the product.
And, the same GPUs are run in datacenter, where there's a big insider risk that vendors want to mitigate with encrypted models (if you're Anthropic and you want AWS to host your models, you don't want 5000 or even 500 or even 50 AWS employees with access to your highly valuable/desirable model weights).
There will be ways for them to segment the market and make their money. Maybe not 100x stock in 5 years but money all the same.
So it is something that has been done for a long time now. Still feels like unethical rent-seeking that should not be tolerated by the customer.
See Also: car companies that charge additional fees for hardware already on the car. Additional services, I sort of understand(remote start via cell network), but stuff like heated seats, that is just wrong.
A license has zero marginal cost in the first place. Software distribution over networks have a negligible marginal cost. An ebook as zero margin cost. We are constantly paying lots of things with zero or negligible marginal cost.
Yep, "Batterygate"
https://www.npr.org/2020/11/18/936268845/apple-agrees-to-pay...
It saves the car manufacturer on logistics cost. The per-unit cost of a seat heater is (much) smaller than the cost related to keeping two variants of seat in stock and to provide spare parts. Say the heating mat costs 100€ per seat including materials and installation, and can be sold to the customer for 400€ permanent, that means as long as 1 in 4 customers buys the seat heating option, the books are balanced, and all money you save on logistics is pure profit.
Adding subscription on top of that is financially even better, if there is one thing companies LOVE thanks to stonk market incentives it is monthly recurring revenue.
I think more than that, it is probably a lie they tell us that you are not paying for the hardware/equipment.
Yeah, Nothing morally wrong there.
AI is a software just like operating systems and executables where open source absolutely dominates the industry. Remember when the compiler engineers optimized the shit out of llama.cpp.
Right now the bottleneck for AI software is compute and data. Moore's law is give us compute, StackOverflow, arXiv (and it's children like bioRxiv), gutenberg and others (3lib, anna, internet.org) will give us data.
It's a waiting game. Open source always wins.
Tons of stuff happenning behind closed doors today though. No more stackoverflow for the next decade, that could create a problem if all new context only exists in chatgpt conversations only openAI has access to (or whoever).
Based on... what? Thoughts and prayers? Feelings? Vibes?
I don't know if you looked around but open source is basically unusable for the largest platform in the world, mobile.
The next platform, whatever that is, will also be closed source (maybe VR, once it finally matures...).
LLMs are closed source, is there even 1 (ONE) LLM which is both open source and open weights and publishes its training data and methodology and is also remotely competitive with the frontier lab LLMs?
Consoles are closed source. Appliances are closed source. Embedded devices are closed source.
Even the web is basically closed source. Blobs of encrypted, obfuscated JS used to serve news pages (!).
Maybe I'm looking in the wrong places, but closed source has won.
And regarding servers, who cares? The platforms users interact with are their clients and the actual server software on top of the Linux OS is closed source 99% of the time.
Don't post lazy comments.
[Hanlon's Razor] This is such a contrived performance drop scenario that I find it hard to believe it was done with the purpose you're inferring.
I'd interpret this as Nvidia not caring that much about Linux desktops/laptops once again (as opposed to Windows desktops/laptops), rather than anything to do with local vs cloud.
Even that argument is weak - the DGX Spark is Linux, and the RTX Spark is practically the same exact chip.
Sorry to remind people of HN, but most of the consumer market still runs Windows. This was a joint project to have a consumer-level Windows-on-ARM machine, likely requiring lots of development efforts from both NVIDIA and Microsoft.
So, for these specific laptops, they put 100% of their attention for it to be on Windows for launch to capture as much market mindshare as they can. Local AI, Graphics and creative work, and some gaming.
I doubt Microsoft would have been happy about NVIDIA having Linux on their laptop right at launch.
Market Share by OS for consumer laptops & desktops:
• Windows: 76.65%
• OS X / macOS (Combined): ~15.67% (OS X at 9.86% + macOS at 5.81% due to user-agent reporting quirks)
• Linux: 5.63%
• Chrome OS: 2.04%
Who exactly?
Beyond him, I would bet most of his bilderberg group friends agree. E.g. Peter Thiel, the owner of JD Vance.
As in, make your own inference and training hardware?
Sovereign AI is less about being able to build it from first principles and more about being able to tinker and modify what already exists. I can’t train a 30B model, but I can modify the weights of one. I can’t fab my own chips, but I can putter with the software that interfaces with them.
Sovereignty is lost when encryption hides your toys and attestation tells you “you may not”.
Estonian only has sovereign AI as long as they’re allowed to buy from Nvidia. Oh wait, they’re already restricted in what they can get.
But those servers most likely use CPUs not made in China, if I were to hazard a guess, doesn't that mean that that AI isn't "truly sovereign" in your mind then?
I stopped using my laptop for dev and switched to talking to claude code on my phone and as you can imagine I am faster at producing code.
100% cloud with a single entry point device to manage dedicated servers + agents
This is a new thing with the firmware that can be used only by Windows. If this NVIDIA policy continues, then NVIDIA will indeed become a bad choice for Linux, but until now, for graphics and AI applications NVIDIA has been the best choice for Linux, by providing the best performance and reliability.
For pure computational applications, especially for those requiring good FP64 throughput, there are many years since NVIDIA has become a worse choice than AMD and more recently also than Intel, but that is due to poor hardware support, not to bad software support.
Hell, their old closed-source NVIDIA _FreeBSD_ drivers were at parity with Windows. CUDA included (though the CUDA toolkit required Linuxulator syscall emulation to run the Linux bins).
Imagine my surprise when my new Blackwell card didn’t work, and that the newly required open source drivers were Linux-only.
If they were using DirectX through Wine, perhaps the DirectX emulation is not optimized for NVIDIA, but only for AMD.
I do not play games, but I have used professional graphics or computational programs using OpenGL/Vulkan/CUDA/OpenCL natively on Linux with the NVIDIA driver, and those have always been as fast or faster than on Windows.
The AMD and the Intel cards have always had poor drivers for OpenGL, which frequently cheated in order to get better benchmark results, by having various rendering defects in comparison with the NVIDIA cards, so when running a benchmark like SPECviewperf, the advantage of NVIDIA has always been much greater than the already higher speed, because that speed was obtained while rendering everything correctly, while with Intel or AMD many things were rendered only approximately.
I actually have more AMD and Intel GPUs than NVIDIA GPUs, but I continue to use the NVIDIA GPUs for graphics, due to a better driver, while I use Intel and AMD for general-purpose computations. However, I have not tested Intel and AMD with a Vulkan graphics application, to see whether they work better with Vulkan than with the obsolete OpenGL.
Those are very different things as far as I understand, and the whole "doesn't work well with linux" really only makes sense on the RTX Spark laptop I think? Isn't the DGX spark linux by default?
> The laptops also support unified memory configurations up to 128GB, shared between CPU and GPU, which is unusual for a laptop and central to why they’re being marketed at AI developers in the first place.
This isn't rare on laptops, pretty default on AMD with iGPU
Also very curious if the NPU is visible in Linux, too.
https://www.xda-developers.com/nvidia-rtx-spark-laptops-same...
It's lack of linux driver support for firmware designed for windows only laptops.
What is going on here? Commenters here are clearly extremely paranoid about NVIDIA centralizing inference. This is not reflected in the facts nor narrative of the article - it's completely fabricated out of left field her in the comments.
This is a problem on the MSI Prestige laptop using the RTX Spark platform with Linux. Lets assume it also applies to the Asus and Dell machines using RTX spark and Linux.
RTX Spark is a new platform and all of these laptops ship with Windows. You cant get Linux on any of these machines from the factory. NVIDIA may even ship drivers for linux later as they did with Dynamic Boost.
I understand that their "secret sauce" might be exposed with open drivers, so I somewhat respect their right to intellectual property in this respect.
A laptop can't run the latest big models, because memory is too expensive. This thing probably can't run models larger than Mistral Small 4. But it'd be a completely different thing to have a hardcoded Mistral Large 4, than a choice of tiny models.
I think I misunderstood what the technology could do a little bit, because I imagined these fixed matrix multiplies encoded into hardware as small-ish, as using routing space etc., not just fast.
I think what I really imagined was useful would be more easily achieved by having a big ROM with model weights connected to an ordinary GPU.
This has been an issue on Linux since 2014, and the gtx980.
Even if this is an intentional nerf by Nvidia, which I doubt, then then nerf is buggy.
Still miles ahead compared to Apple, where linux does not even boot on newer macs!
If you really want to do DIY inference, plug it in at home and set up tailscale so you can access it on the go.
No huge battery drain or toasty hot laptop required. Lower cost laptop. Lighter laptop. Longer battery life laptop. Win win win win.
Maybe you need a therapy. FOMO can be treated, too. Or you're just upset about Trump falling into irrelevance.
The market for someone who wants local AI in a laptop for $10k has to be tiny.
To me the simple answer to why is control. If I run a model locally I get to control the model params, the system prompt, all of it. And of course there is no quota. But probably most importantly I am not tied to paying some company a monthly subscription forever.
I meanwhile, run my ridiculously OP workstation class hardware as a server - 9995wx, 768GiB, rtx pro 6000 w/ 96GiB vram - at home, and my Asus ZenBook is light, cool, and long-lived.
Runs GLM 5.3 flash at 35 tok/sec while consuming only 15 GiB VRAM, and I can run a lot more stuff on there alongside.
unique local data and know how not shared with cloud.
GGP's point is just: keep the bulk of your compute at home, where airflow and energy is cheaper (in an opportunity-cost sense). Keep the laptop portable, cool, and light, with long battery life by offloading your own computing tasks to your own hardware.
I am not saying don't use a laptop.
I am not saying you have to use cloud inference.
I am saying using a laptop for inference is dumb. If you want private inference:
- use a dgx spark or similar plugged in at home
- set up tailscale so you can connect from anywhere
- use your phone as a hotspot on the go
- work on your laptop, using tailscale to connect to your home inference server if you're on the go
This gets you everything except for the few hours you might be on an airplane.
The simple point I'm always trying to make is that area vs volume is a fundamental physical difference regarding cooling, which is critical for high performance applications (that's why they're considering an RTX Spark in the first place, right?).
There's so much more I could say about my Strix Halo machine, which fits in the palm of my hand and I have a wonderful portable screen and keyboard+touchpad (and only cost me 2k with 128 GB fast memory, idles at 4w, amazing 32 thread Zen 5 CPU with AVX-512, ...), but it's either preaching to the choir or pissing against the wind.
But with Nvidia people just expect they nerf all the features.
If you want to prosper with the current of nature, provide a thing easily for mutual benefit. The more you try to block a thing, the sooner your demise into uselessness.
Absolutely hilarious.
This is just incompetence. No reason Nvidia with its TRILLIONS of market cap couldn't of tested and fixed this on Linux before shipping.
Or a day one patch.
I have no doubt that'll it'll work right eventually. Long after RTX Spark laptops flop and can be purchased for half off.
Last night I pulled out my 5070TI laptop and installed CachyOS. I was running various models within about 20 minutes.
That 12GB of VRAM is a hard limitation, Unsloth can't split jobs between vram and system memory.
Could have tested.
Even from the article I'm guessing you could write a script to automatically extract the encryption keys while on Windows and the load those in Linux.
This is only mildly inconvenient for anyone who'd use this in a data center cluster.
Why the fuck is the power table encrypted?
> laptop ran a full desktop at 4K and 120Hz, with working audio, touchpad, keyboard, and CUDA acceleration
is amazing. It seems that this person's MSI version had some quirks with firmware but Omarchy already supports the Asus and Dell versions of the RTX Spark out of the box. I'm surprised that everyone is focusing on the negative when, prior to RTX Spark's release, it wasn't clear if Linux would work on the RTX Spark at all in the first place.
This is way better support than Qualcomm's Snapdragon X1 and X2 laptops, which are still struggling to get Linux support years after release.
Presumably, once this one power management driver issue gets sorted out (surely an accidental bug rather than what some of y'all cynics are suggesting), it will be great. Now if only it could be cheaper...
NVIDIA and Microsoft tried so hard to make a Mac clone with this and it’s just missing the mark everywhere.
It doesn’t do the dynamic memory split that a Mac does so out of the box it reserves 32GB (half) for the GPU and windows limits CPU processes to that. Getting more memory requires a reboot.
And because I want Unix tools (without Cygwin), WSL2 splits that even further in half.
Performance is way behind a similarly specced Mac. The GPU is thermally throttled near a 5060 laptop. The CPU out of the box came in a lower power state but windows power management makes it stuttery unless you change that, at which point it’s cooling is a hurricane compared to the Mac which outperforms it while being quieter.
The new MagSafe like connector is stupidly designed too. You can’t generically use it in any USB-C port, it’s super chonky and disconnects only on the horizontal axis. Worse, because it’s a right angle connector it can only be used in one orientation without blocking the other ports.
Windows input handling sucks too. There’s a lag before the trackpad responds based on moved distance. Some terrible take on palm rejection I’m sure. But even their display is mediocre because it gets very bright but they messed up local dimming. So if you select a bunch of text in the terminal it blinds you.
Idk I got this specifically to do windows dev with CUDA on arm. So I knew it was niche but anyone else buying this is making a mistake. It’s an unfinished product trying to be a competitor that nobody in Microsoft even understands.
This Linux support issue is just the icing because everyone keeps asking me why I don’t just put Linux on it and workaround my issues, but you can’t.
I’ve spoken to friends at NVIDIA who say it’s rushed out. So maybe software will get better but I urge folks to not buy this unless you’re into the same niche needs as me.
Odd because Tegra/Jetson platforms have been doing this for years.
I suspect it’s a limitation of them focusing on Windows driver and windows being far behind the curve.
But, on the other hand, having my DGX Spark stack at hand, I know the solution:
"AI views proprietary components as damage and routes around them."Sovereign AI has never been more important.
But yeah, we do need to make sure the REA cat stays out of the bag.
I, for one, welcome our reverse-engineer-anything overlords.
TIL. Nice word.
EDIT: I don't know why you changed your comment, but yes, it was obvious from reading the article that it was not actually about the DGX Spark.
Or maybe just not at all - shell out for OAI tokens and send us all your data - your privacy is very valuable to us and we will make sure to be the only ones to monetize it.
You could say, Nvidia is selling pickaxes and has no business gatekeeping what rocks you hit with it, but I guess it depends on their relationships with major mining companies, and ultimately what pays more.
Vote with Your wallet.
Its that simple.
Every.
Single.
Time.
They've always played these sorts of games. Now, people are seeing the substandard solutions for what it is, and are angry.
They really got it right on multiple levels - for those who don't know, roller coaster tycoon was infamously written in 99% assembly and meticulously optimized by the developer to run effectively in 16 megabytes of ram.
We can see with this evil cartel currently driving up the RAM prices how greedy they all are. Our money should go into fairness. The US court system has given up on that. Even on capitalism. Capitalism used to claim that a free market with competition leads to lower prices. Well - we have several de-facto cartels here. And nobody is breaking any of these up.