I don't think there's any incentive for Nvidia to make this a Windows-only device, so most likely it will be fully supported on Linux, just like their GPUs are.
What trouble? If you want a GPU that works on Linux, let alone FreeBSD, you buy nVidia, install their drivers and get on with your life (and sure, maybe you can't use Wayland, but why would you want to?). I'm all for open-source in theory, but in practice the AMD drivers cause far more trouble than the nVidia ones ever do.
Depends. It is the typical Nvidia problem. Everything is a black box but when it all works it is the best option available. But when it breaks, you hate them with a passion.
Sorry, but when it comes to chipsets, they're not even close.
The DGX Spark uses a GB10 with 128 GB unified LPDDR5X memory, while the DGX Station has a GB300 with 496 GB LPDDR5X (CPU) + 252 GB HBM3e (GPU) memory.
It's like Little League versus Major League, which is why the latter costs about 20 times more than the former.
The fact that both run Linux is just because they're part of the same DGX family.
Doesn't it come with Nvidia's blend of Ubuntu with a custom kernel? Do other distros work as well as "DGX OS" or are nvidia's kernel changes pretty important to have?
This plus the price different had me buy an AMD Strix Halo board last week. It seems the work with vLLM and training models could make the Spark worth the price difference, but before today's news I had the same thought about support and did not want to lock myself into a cool paperweight, especially with 128gb of RAM on the line. AMD is x86 and I can repurpose that or run Linux forever.
I've not noticed much in it that is NVIDIA specific.
But I would say that as an Ubuntu and Debian user for decades I have no incentive to use anything else on it and I'm just pleased to have a Linux on Aarch64 machine that is well supported for a change.
afaict, they have their own package repo mirrors and a few dedicated packages for nvidia stuff
tbh, I was rather unimpressed with the out-of-box experience for an "ai" computer, you couldn't even run a model locally with the common tools people use (no llama-cpp, ollama, vllm, etc). No huggingface CLI eiher, like come on!
I need to update that because I have a nice vllm setup on there now with 4 models running, but should be able to get anyone else going without having to muddle about as I did.
Honestly this looks like Microsoft must have thrown a pile of money at them to not mention it, as it's just too obviously the main question.
No one seriously cares about this running Windows. We want Steam and CUDA/Ollama, and Windows just gets in the way. nVidia are simply not that oblivious, but I have to admit in their position I'd have considered the Microsoft involvement more trouble than it's worth, which is among the many reasons I'm not a billionaire.
Maybe they think the RAM market is so terrible it will kill the whole initiative regardless.
I’ve read all the stuff about how llama.cpp is much faster and better than ollama, and i believe it - but good god llama.cpp isn’t user friendly.
You’d think in an era where “code is free” there would be an easier story around running local ai than compiling llama.cpp by hand and then spending hours researching flags - only for it to crash from an oom error every ten prompts or so.
You're supposed to use a cheap ChatGPT subscription to run optimization loops over llama.cpp flags with a self-contained reproducible benchmark script and just let it burn for hours/days until it is fully optimized ))))
Valve did that little more than a decade ago, the original Steam Machines. It didn't take, and despite the success of the Deck and current techy trends, Linux does not have the % to make the ROI worthwhile if it isn't simple for developers. Proton is a wedge in the door that will help Linux get there.
A potential change in Valve's culture/management aside, "let valve do the work" is a feature, not a bug. Studio spends all their budget targeting one platform (which still has ~90+% of the PC gaming market), and get Linux support for free.
Windows' monopoly on game dev isn't just market share either, since game dev isn't just code. You still need Photoshop, Maya, etc. and in smaller studies there's typically a crossover where some devs are doing art as well. Visual Studio's C++ debugger is still one of the best, and the tooling elsewhere hasn't caught up yet (compared to DX + PIX).
Then you also have to solve distribution and handling the fragmented display & audio stack. It's gotten a lot better, but its still a factor.
I'm fine with most of the work going into Wine/Proton. A stable ABI for Linux is a boon, if it happens to be Win32 then so be it.
And yet where are the native games for the platform?
Note how game developers rather spend their working hours in Windows with all its issues, even if they happen to have Linux servers for running their MMOs.
... after you watch these ads, or pay for our premium subscription. After all, those games aren't going to host themselves and your license doesn't allow an alternative.
In truth if AMD or nVidia put their mind to having decent profiling tooling on Linux, and the AI wave suggests they will have no option, then this could readily become a thing.
There are two new things being announced here: the GB10 chip being put into laptops, and GB10 running Windows. GB10 running Linux is not news, it's a product that's been shipping since last fall.
I don't think there's any incentive for Nvidia to make this a Windows-only device, so most likely it will be fully supported on Linux, just like their GPUs are.
There's also the precedent of the several ARM Linux systems they shipped that tended to have much worse support.
> just like their GPUs are
So with proprietary blobs that give you more trouble that they're worth?
Those blobs are worth $5T; show some respect.
Kids these days, amirite?
What trouble? If you want a GPU that works on Linux, let alone FreeBSD, you buy nVidia, install their drivers and get on with your life (and sure, maybe you can't use Wayland, but why would you want to?). I'm all for open-source in theory, but in practice the AMD drivers cause far more trouble than the nVidia ones ever do.
Depends. It is the typical Nvidia problem. Everything is a black box but when it all works it is the best option available. But when it breaks, you hate them with a passion.
I've never had a single problem with my Nvidia GPUs on Linux.
They worked with Microsoft to make this commercially viable. That’s possibly reason enough.
I wouldn't trust it to have good upstream support. It's Nvidia. So not really interested.
An nvidia engineer in a discord server said it should work fine
Sort of. It's the same chipset as in the DGX Spark & DGX Station, which run Ubuntu (NVIDIA's flavor).
Sorry, but when it comes to chipsets, they're not even close. The DGX Spark uses a GB10 with 128 GB unified LPDDR5X memory, while the DGX Station has a GB300 with 496 GB LPDDR5X (CPU) + 252 GB HBM3e (GPU) memory. It's like Little League versus Major League, which is why the latter costs about 20 times more than the former. The fact that both run Linux is just because they're part of the same DGX family.
I took "same" to mean "compatible driver and software stack" not "compute perf", rightly or wrongly.
DGX Spark comes with linux out of the box, it would be hard to imagine this device is not also compatible
Doesn't it come with Nvidia's blend of Ubuntu with a custom kernel? Do other distros work as well as "DGX OS" or are nvidia's kernel changes pretty important to have?
Hopefully better than support on their Jetson or orin boards, where compiling anything is hard because of the outdated stack.
This plus the price different had me buy an AMD Strix Halo board last week. It seems the work with vLLM and training models could make the Spark worth the price difference, but before today's news I had the same thought about support and did not want to lock myself into a cool paperweight, especially with 128gb of RAM on the line. AMD is x86 and I can repurpose that or run Linux forever.
I've not noticed much in it that is NVIDIA specific.
But I would say that as an Ubuntu and Debian user for decades I have no incentive to use anything else on it and I'm just pleased to have a Linux on Aarch64 machine that is well supported for a change.
For some value of "well supported" - NVIDIA's own internal catalogs (libraries, NIMs, etc) are still spotty on aarch64 coverage.
afaict, they have their own package repo mirrors and a few dedicated packages for nvidia stuff
tbh, I was rather unimpressed with the out-of-box experience for an "ai" computer, you couldn't even run a model locally with the common tools people use (no llama-cpp, ollama, vllm, etc). No huggingface CLI eiher, like come on!
I did put together my eventual setup in a repo https://github.com/verdverm/sparky
I need to update that because I have a nice vllm setup on there now with 4 models running, but should be able to get anyone else going without having to muddle about as I did.
Honestly this looks like Microsoft must have thrown a pile of money at them to not mention it, as it's just too obviously the main question.
No one seriously cares about this running Windows. We want Steam and CUDA/Ollama, and Windows just gets in the way. nVidia are simply not that oblivious, but I have to admit in their position I'd have considered the Microsoft involvement more trouble than it's worth, which is among the many reasons I'm not a billionaire.
Maybe they think the RAM market is so terrible it will kill the whole initiative regardless.
You misspelled llama.cpp
I’ve read all the stuff about how llama.cpp is much faster and better than ollama, and i believe it - but good god llama.cpp isn’t user friendly.
You’d think in an era where “code is free” there would be an easier story around running local ai than compiling llama.cpp by hand and then spending hours researching flags - only for it to crash from an oom error every ten prompts or so.
You're supposed to use a cheap ChatGPT subscription to run optimization loops over llama.cpp flags with a self-contained reproducible benchmark script and just let it burn for hours/days until it is fully optimized ))))
WSL is the answer in what most folks are concerned.
Has Steam finally started to push for native Linux games instead of translating Windows ones?
Valve did that little more than a decade ago, the original Steam Machines. It didn't take, and despite the success of the Deck and current techy trends, Linux does not have the % to make the ROI worthwhile if it isn't simple for developers. Proton is a wedge in the door that will help Linux get there.
It is simple, Android NDK has all the same APIs for 3D rendering and audio, as do all major middleware engines.
The failure of business, only reinforces Windows as the platform most studios reach for.
Buy Windows, buy Visual Studio, pay game engines licenses, let Valve do the work.
This ignoring that current Valve's management doesn't live forever, so who knows what happens afterwards.
A potential change in Valve's culture/management aside, "let valve do the work" is a feature, not a bug. Studio spends all their budget targeting one platform (which still has ~90+% of the PC gaming market), and get Linux support for free.
Windows' monopoly on game dev isn't just market share either, since game dev isn't just code. You still need Photoshop, Maya, etc. and in smaller studies there's typically a crossover where some devs are doing art as well. Visual Studio's C++ debugger is still one of the best, and the tooling elsewhere hasn't caught up yet (compared to DX + PIX).
Then you also have to solve distribution and handling the fragmented display & audio stack. It's gotten a lot better, but its still a factor.
I'm fine with most of the work going into Wine/Proton. A stable ABI for Linux is a boon, if it happens to be Win32 then so be it.
At this point Valve look more capable of running a platform business than Microsoft do.
Microsoft have spent the whole Nadella era in "oooo cloud" inspired wonder and actively screwed up everything else.
And yet where are the native games for the platform?
Note how game developers rather spend their working hours in Windows with all its issues, even if they happen to have Linux servers for running their MMOs.
> Valve's management doesn't live forever, so who knows what happens afterwards.
Tens of thousands of Windows games would remain playable with ubiquitous Vulkan-capable hardware and a 500mb Proton runtime?
... after you watch these ads, or pay for our premium subscription. After all, those games aren't going to host themselves and your license doesn't allow an alternative.
If it runs faster than the windows ones, who cares?
The game developers that use Windows, with Visual Studio, to develop such games.
This is, admittedly, the great anomaly.
In truth if AMD or nVidia put their mind to having decent profiling tooling on Linux, and the AI wave suggests they will have no option, then this could readily become a thing.
This is strangely absent from the news.
There are two new things being announced here: the GB10 chip being put into laptops, and GB10 running Windows. GB10 running Linux is not news, it's a product that's been shipping since last fall.
It's a collaboration with Microsoft so going to say no, probably not.