AMD Releases ROCm 5.7 GPU Compute Stack

[phoronix]

Phoronix: AMD Releases ROCm 5.7 GPU Compute Stack

ROCm 5.7 was released on Friday with the introduction of a new "hipTensor" library, the ROCgdb debugger being extended with Fortran and OMPD support, and new optimizations to the rocRAND and MIVisionX libraries. AMD has also announced end-of-support for the AMD Instinct MI50 accelerator while not yet formally announcing any new RDNA3 GPU support...

https://www.phoronix.com/news/AMD-ROCm-5.7-Released

[LumielGR]

For Arch Linux I've updated opencl-amd and opencl-amd-dev packages to 5.7 version. I've also included hipTensor that looks to be missing from the Ubuntu release. But I believe only CDNA is supported.

[david-nk]

At this point AMD should consider serious changes, including firing every single person in the decision making chain that is responsible for there not being day 1 support for new graphics cards.

I already bit the bullet and got a RTX 4090 despite at least the AMD graphics side of the drivers being vastly superior to Nvidia's, but no ML support is obviously a no-go.

I'll check back to reassess the situation in 5 years or so, but I wouldn't be surprised if the AMD situation still hasn't improved by then.

[xpris]

Anyone know if is still possible to install/tun ROCm 5.7 on Polaris GPU (500 series)?

[Beach]

This is so incredibly disappointing.

I really want to support AMD, but they make it very hard to do so.

[timofonic]

This is so incredibly disappointing.

I really want to support AMD, but they make it very hard to do so.

I agree.

RIP ROCm and AMD for GPU computing tasks.

Nvidia is very happy about this.

AMD should be renamed to Fake Promises Inc.

bridgman


What do you say about this?

[osw89]

I already bit the bullet and got a RTX 4090 despite at least the AMD graphics side of the drivers being vastly superior to Nvidia's, but no ML support is obviously a no-go.

ML support isn't missing, official support is. You can still use ROCm if you happen to have something that uses the same chips as one of the few officially supported workstation cards. I have been using ROCm with my RX6800 for months and haven't really had any problems. Don't get me wrong the situation is still atrocious but not as bad as people seem to think. It works on ~15 cards instead of 3 .

[Zyten]

I also feel quite disappointed in the support of GPUs in ROCm. I‘ll probably have access to a RX 7700 XT soon and I‘ll test if and what parts of ROCm actually work, but with the current official support AMD is miles behind offers from NVIDIA.

I really hope that they’ll catch up soon. I mean, some
parts of ROCm seem to support RX 7000 for some time now. However, the documentation for that is extremely sparse and most information has to be gathered from external sources.

[Lycanthropist]

PyTorch Nightly works fine here with ROCm 5.6 on a 7900 XTX. E.g. Stable Diffusion inference and training.

[darkbasic]

Hopefully ROCm 6 will start working on non-x86 architectures as well: https://github.com/RadeonOpenCompute...ime/issues/158