Intel’s Arc GPUs will compete with GeForce and Radeon in early 2022

Two weeks ago, the Hardware Survey reported Linux breaching 1% for the first time ever [1], for reasons not related to Deck (in fact, its not obvious WHY linux has been growing; disappointment in the Win11 announcement may have caused it, but in short, its healthy, natural, long-term growth). I would put real money up that Linux will hit 2% by the January 2022 survey, and 5% by January 2023.

Proton short-circuits the porting argument. It works fantastically for most games, with zero effort from the devs.

We're not talking about Linux being the majority. But its definitely looking like it will see growth over the next decade.

[1] https://www.tomshardware.com/news/steam-survey-linux-1-perce...

On my end, and my Year of Linux Desktop started almost a decade ago, the Linux experience has not only never been better (of course) but has improved faster in recent years than ever before, and gaming is one of the fastest improving areas.

Proton is a fantastic piece of software.

https://www.pcgamer.com/nvidia-ampere-samsung-8nm-process/

This is one of the smartest moves by Intel, make their own stuff and consume production from all their competitors, which do nothing but paper designs. Nvidia and especially AMD took a risk not being in the fabrication business, and now we'll see the full repercussions. It's a good play (outsourcing) in good times, not so much when things get tight like today.

Disclosure: Work at Google, but not on Stadia.

With that said, I don't know if it's a sign of the shortage coming to an end; I think the release of the Ryzen 5700G with integrated graphics likely helped bridge the gap for people who wanted low-end graphics without paying the crazy markups.

The issue is that AMD has split their compute into two categories:

* RDNA -- consumer cards. A new ISA with new compilers / everything. I don't think its reasonable to expect AMD's compilers to work on RDNA, when such large changes have been made to the architecture. (32-wide instead of 64-wide. 1024 registers. Etc. etc.)

* CDNA -- based off of Vega's ISA. Despite being "legacy ISA", its pretty modern in terms of capabilities. MI100 is competitive against the A100. CDNA is likely going to run Frontier and El Capitan supercomputers.

------------

ROCm focused on CDNA. They've had compilers emit RDNA code, but its not "official" and still buggy. But if you went for CDNA, that HIP / ROCm stuff works enough for the Oak Ridge National Labs.

Yeah, CDNA is expensive ($5k for MI50 / Radeon VII, and $9k for MI100). But that's the price of full-speed scientific-oriented double-precision floating point GPUs these days.

They had no choice. Getting a bunch of HPC people to completely rewrite their code for a different API is a tough pill to swallow when you're trying to win supercomputer contracts. Would they have preferred to spend development resources elsewhere? Probably, they've even got their own standards and SDKs from days past.

> everyone else using GPU's is running fast as they can towards Vulkan

I'm not qualified to comment on the entirety of it, but I can say that basically no claim in this statement is true:

1. Not everyone doing compute is using GPUs. Companies are increasingly designing and releasing their own custom hardware (TPUs, IPUs, NPUs, etc.)

2. Not everyone using GPUs is cares about Vulkan. Certainly many folks doing graphics stuff don't, and DirectX is as healthy as ever. There have been bits and pieces of work around Vulkan compute for mobile ML model deployment, but it's a tiny niche and doesn't involve discrete GPUs at all.

> Is it just too soon too early in the adoption curve

Yes. Vulkan compute is still missing many of the niceties of more developed compute APIs. Tooling is one big part of that: writing shaders using GLSL is a pretty big step down from using whatever language you were using before (C++, Fortran, Python, etc).

> do ya'll think there are more serious obstructions long term to building a more Vulkan centric AI/ML toolkit

You could probably write a whole page about this, but TL;DR yes. It would take at least as much effort as AMD and Intel put into their respective compute stacks to get Vulkan ML anywhere near ready for prime time. You need to have inference, training, cross-device communication, headless GPU usage, reasonably wide compatibility, not garbage performance, framework integration, passable tooling and more.

Sure these are all feasible, but who has the incentive to put in the time to do it? The big 3 vendors have their supercomputer contracts already, so all they need to do is keep maintaining their 1st-party compute stacks. Interop also requires going through Khronos, which is its own political quagmire when it comes to standardization. Nvidia already managed to obstruct OpenCL into obscurity, why would they do anything different here? Downstream libraries have also poured untold millions into existing compute stacks, OR rely on the vendors to implement that functionality for them. This is before we even get into custom hardware like TPUs that don't behave like a GPU at all.

So in short, there is little inevitable about this at all. The reason people may have been frustrated by your comment is because Vulkan compute comes up all the time as some silver bullet that will save us from the walled gardens of CUDA and co (especially for ML, arguably the most complex and expensive subdomain of them all). We'd all like it to come true, but until all of the aforementioned points are addressed this will remain primarily in pipe dream territory.

Aurora was originally slated to use Phi chips, which are an unrelated architecture to these GPUs. The delays there don't say much about problems actually getting this new architecture out. It's more that they were halfway through making a supercomputer and then started over.

I could probably pin the biggest share of the blame on 10nm problems, which are irrelevant to this architecture.

As far as this architecture goes, when they announced Aurora was switching, they announced 2021. That schedule, looking four years out for a new architecture, has only had one delay of an extra 6 months.

> In the lead-up to the i740's introduction, the press widely commented that it would drive all of the smaller vendors from the market. As the introduction approached, rumors of poor performance started circulating. … The i740 was released in February 1998, at $34.50 in large quantities.

However, this suggests that it was never going to be a top-end contender since it was engineered to hit a lower price point and was significantly under-specced compared to the competitors which were already on the market:

> The i740 was clocked at 66Mhz and had 2-8MB of VRAM; significantly less than its competitors which had 8-32MB of VRAM, allowing the card to be sold at a low price. The small amount of VRAM meant that it was only used as a frame buffer, hence it used the AGP interface to access the system's main memory to store textures; this was a fatal flaw that took away memory bandwidth and capacity from the CPU, reducing its performance, while also making the card slower since it had to go through the AGP interface to access the main memory which was slower than its VRAM.