ROCm 5.5.1 Released For AMD's Open-Source Compute Stack

[wsippel]

I'm still hoping that AMD will support consumer grade cards in the future. It is a pitty that you can't just check out rocm with your already purchased card. Or just as hobbiest. This could help to underwhelm Nvidias CUDA domination.

AMD lists cards, but the actual models don't really matter. ROCm is built for GPU architectures, and therefore works on every card of a given architecture. It's currently built for the following targets:

• gfx900: Vega 64
• gfx906: Radeon VII
• gfx908: Instinct MI100
• gfx90a: Instinct MI200/ MI250
• gfx1030: RX 6800/ 6900/ 6950, also works on many other RDNA2 and RDNA1 cards with the environment variable "HSA_OVERRIDE_GFX_VERSION=10.3.0"
• gfx1100: RX 7900
• gfx1101: Unreleased, probably RX 7700/ 7800?
• gfx1102: RX 7600

Only the gfxXXXX string really matters, the "support matrix" is about tech support, not hardware support.

[Quackdoc]

I am dismayed to hear that ROCm is required to use OpenCL (and I think OpenMP for GPGPU), and that it doesn't support consumer cards. There's so much compute resources bottled up in the latest cards that we should be able to take advantage of.

thankfully rusticl is starting to shape up quite well. ive never touched openMP nor know anyone who has so I cant comment there

[agd5f]

agd5f Frontier uses rocm? Be it upstream or modified versions of it

Same version you can get from AMD.

[CochainComplex]

AMD lists cards, but the actual models don't really matter. ROCm is built for GPU architectures, and therefore works on every card of a given architecture. It's currently built for the following targets:

• gfx900: Vega 64
• gfx906: Radeon VII
• gfx908: Instinct MI100
• gfx90a: Instinct MI200/ MI250
• gfx1030: RX 6800/ 6900/ 6950, also works on many other RDNA2 and RDNA1 cards with the environment variable "HSA_OVERRIDE_GFX_VERSION=10.3.0"
• gfx1100: RX 7900
• gfx1101: Unreleased, probably RX 7700/ 7800?
• gfx1102: RX 7600

Only the gfxXXXX string really matters, the "support matrix" is about tech support, not hardware support.

Thx for the list. I was already using the HSA_OVERRIDE_GFX_VERSION prefix but that is not the full story - see comments:

https://www.reddit.com/r/StableDiffusion/comments/ww436j/howto_stable_diffusion_on_an_amd_gpu/

[Sanjuro]

Is it possible to install ROCm on Slackware or does it have to be Ubuntu/Redhat?
Can it run on latest kernels (6.1/6.3)?
Does it need some custom patches for kernel/llvm/mesa etc.?

I want to buy new gpu but I'm afraid to get AMD again. My current gpu (rx480) has given me a lot of trouble with blender. And now it's basically useless because of dropped opencl support.

I want to play with StableDiffusion, Blender, UE5 and games. I need something with at least 12GB vram so I was thinking about 6700xt/6800.
AMD cards work nice with linux desktop and linux gaming so it's hard to decide.

Or should I just get rtx 3060 12GB?

[bridgman]

I am dismayed to hear that ROCm is required to use OpenCL (and I think OpenMP for GPGPU), and that it doesn't support consumer cards. There's so much compute resources bottled up in the latest cards that we should be able to take advantage of.

We have been supporting the ROCm stack up to and including OpenCL on consumer hardware for a couple of years - it's the higher level components, particularly the math libraries (which use a lot of shader assembly code) that have required more effort and hence more time.

We don't make that as clear in the documentation as I would like. Will see if we can improve that.

[Developer12]

agd5f Frontier uses rocm? Be it upstream or modified versions of it

Until I can run ROCm reliably on the consumer card in my desktop without having to play stupid games, then I flatly refuse to develop for it. Simple as that.

The lost market share to nvidia will continue until support improves.

EDIT: I specifically need HIP. nobody writes code for openCL anymore.

[boxerab]

image.png

Until rocm is available for all consumer cards on both Linux and Windows, NVIDIA will continue to dominate the market.

[darkbasic]

I am dismayed to hear that ROCm is required to use OpenCL

Nope, you can use Rusticl.

[stompcrash]

We have been supporting the ROCm stack up to and including OpenCL on consumer hardware for a couple of years - it's the higher level components, particularly the math libraries (which use a lot of shader assembly code) that have required more effort and hence more time.

We don't make that as clear in the documentation as I would like. Will see if we can improve that.

Hey, that's great to hear. I've got to try it out now.