AMD Releases ROCm 4.2 Compute Stack

[cb88]

So, does it work on NAVI or not? I still recall 2019, when I was jokingly saying that it would take a lot of time for AMD to support ROCm on NAVI. Boy it's taking forever

ROCm's OpenCL works... not all of HIP and the roc Libaries work yet which are important for it being a CUDA replacement.

It's taking forever because it is a LOT more than just a compiler these days.

[extremesquared]

It's becoming a bit of a joke, an entire software stack that only supports 3 Vega card models that no one owns.
I wish I could use an AMD card for ML work since Nvidia's drivers are just a buggy mess, but apparently that won't happen anytime soon.

I am really trying not to be the stereotypical entitled GPU consumer, but the timeline here is getting a little bit ridiculous. There are quite a few workstation cards in the wild at this point without practical OpenCL support.

[bridgman]

I am really trying not to be the stereotypical entitled GPU consumer, but the timeline here is getting a little bit ridiculous. There are quite a few workstation cards in the wild at this point without practical OpenCL support.

Which workstation cards are you talking about ? The ROCm stack has been running on Navi up to OpenCL in the Linux packaged drivers since 20.45, and before that we shipped PAL-based OpenCL for Navi.

[extremesquared]

Which workstation cards are you talking about ? The ROCm stack has been running on Navi up to OpenCL in the Linux packaged drivers since 20.45, and before that we shipped PAL-based OpenCL for Navi.

That is very interesting. What is keeping them off the official support list? I've avoided attempts due to the somewhat haphazard management of the github issue tracker -- where all navi+ issues still seem to just get closed without resolution.

My primary (most modern) card is an amd pro w5500.

[Mathias]

I think someone should put together an awesome-AMD-compute page with all the stuff you *can* do on AMD cards. Apart from Blender, Hashcat and Mining, there is not much I know of that can really use OpenCL in the wild. Darktable has OpenCL support, but I never found a filter that is faster on my GPU than on my 8-Core Ryzen. Also processing takes only a few seconds anyways.

[MadeUpName]

I think someone should put together an awesome-AMD-compute page with all the stuff you *can* do on AMD cards. Apart from Blender, Hashcat and Mining, there is not much I know of that can really use OpenCL in the wild. Darktable has OpenCL support, but I never found a filter that is faster on my GPU than on my 8-Core Ryzen. Also processing takes only a few seconds anyways.

Resolve can use openCL on AMD. It just can't be ROCM, it has to be AMDGPU-Pro which is only available for 2 distros. And it can't be any version other than 1.2 as AMD broke some thing when they went past that.

[qarium]

Which workstation cards are you talking about ? The ROCm stack has been running on Navi up to OpenCL in the Linux packaged drivers since 20.45, and before that we shipped PAL-based OpenCL for Navi.

can you tell me something about the new people amd is hireing...? is it possible to hire even more people ?

i am on a vega64 and 1920x i want to upgrade to 2950x and a 6800XT or 6900XT but man i wish amd could do a better job on ROCm...
yes i unterstand amd does enterprise first to make money and i have no problem with that
i would even buy a 12nm backport of the RDNA architectur...

i think the time of money shortage is over for -AMD so they could do much more.

[wizard69]

can you tell me something about the new people amd is hireing...? is it possible to hire even more people ?

I have not idea what problems AMD might be having but I do know that finding qualified workers in a different tech industry is extremely difficult.

i am on a vega64 and 1920x i want to upgrade to 2950x and a 6800XT or 6900XT but man i wish amd could do a better job on ROCm...
yes i unterstand amd does enterprise first to make money and i have no problem with that
i would even buy a 12nm backport of the RDNA architectur...

It isn't so much enterprise but rather government contracts for super computing that has AMD focuses on CDNA. This is so important to AMD that I really doubt consumer cards will get much focus at all this year. Once ROCm is good enough to meet contract obligations they will likely expand hardware support.

i think the time of money shortage is over for -AMD so they could do much more.

It isn't that easy. You can't hire stupid people to make advancements in such a project. By the way I don't use the word "stupid" here lightly, people seem to forget that you need people with innate capability to pull off these projects. Further even a "smart" person takes time to get up to speed, hire a person today and it cold be 6 months later before they are contributing significantly.

[vegabook]

I got tired of waiting: https://www.ebay.co.uk/itm/124699576...8AAOSwfBhgiIof

Flogged the VII in which I had invested so many FP64 and machine learning hopes and dreams.

Koduri left AMD with this card in development, it was far enough along that they actually launched it. But it was "politically incorrect" 'cos it was too powerful on compute and not powerful enough on gaming, and was a Koduri creation, so AMD ignored it. Trashed their own creation. Within weeks almost of its launch AMD was already allowing rumours to fly that it was a dead end.

AMD screwed its customers on Radeon VII, the best card they've ever produced. I will NEVER buy AMD graphics again.

Thank you Raja Koduri (no, not you, AMD, and your latter day 192-bit cynicism), for the Radeon VII's ethermining aftermarket value.

[vegabook]

I have not idea what problems AMD might be having but I do know that finding qualified workers in a different tech industry is extremely difficult.

It isn't so much enterprise but rather government contracts for super computing that has AMD focuses on CDNA. This is so important to AMD that I really doubt consumer cards will get much focus at all this year. Once ROCm is good enough to meet contract obligations they will likely expand hardware support.


It isn't that easy. You can't hire stupid people to make advancements in such a project. By the way I don't use the word "stupid" here lightly, people seem to forget that you need people with innate capability to pull off these projects. Further even a "smart" person takes time to get up to speed, hire a person today and it cold be 6 months later before they are contributing significantly.

How much does it cost to hire 20 top people from Nvidia (or Intel)? Ultra generously, 500k each per year? That's 10 million dollars per year. Since end of 2018 AMD has massive fundraising capacity thanks to its market capitalisation rocketing above 20 billion that's 20, thousand, million (and today, 100 billion). AMD is ENTIRELY capable of having made this happen 2.5 years ago, more than enough time for the skills ramp-up, especially if you pay properly and hire properly, consistent with the fact that as AMD, at the time, is one of only 2 entities on earth with credible GPU IP. Yet here we are in mid 2021, and AMD is still flailing about on compute while Nvidia laughs all day long.

This ROCm mess has absolutely no excuses. Ever since AMD bought ATI, RTG has been a second class citizen. Mercy me this division still has value and needs to be SOLD out of AMD's incompetent (GPU-wise) hands sharphish before Intel (or Qualcomm) eat the leftovers of the lunch that Nvidia has already and continues to feast on.

I get Cuda 10 on a 100 dollar Jetson Nano. For C**** sake. Don't tell me a 700 dollar Navi card is still waiting. Honestly.

Challenge to AMD: fire the entire ROCm management team and watch your cap go up 20%.