AMD Releases ROCm 4.2 Compute Stack

[phoronix]

Phoronix: AMD Releases ROCm 4.2 Compute Stack

Just over one month has passed since the release of Radeon ROCm 4.1 and it's now been succeeded by the ROCm 4.2 feature release...

https://www.phoronix.com/scan.php?page=news_item&px=AMD-ROCm-4.2-Compute

[bridgman]

Best link for release notes etc... is the top-level repo: https://github.com/RadeonOpenCompute/ROCm

[cb88]

At least AMD listened and opened Radeon Rays again... hopefully they are listening to the people needing ROCm on RNDA.

[extremesquared]

Radeon
RX 550
(Lexa)[22][23] April 20, 2017

Looks like we just missed the fourth birthday of the most recent consumer GPU to be officially supported by ROCm. They grow up so fast.

EDIT: I am wrong. Three-year-old vega is supported.

[coder]

Looks like we just missed the fourth birthday of the most recent consumer GPU to be officially supported by ROCm. They grow up so fast.

EDIT: I am wrong. Three-year-old vega is supported.

Radeon VII launched a little over 2 years ago.

[tehehe]

Radeon VII launched a little over 2 years ago.

And doesn't work with any of rocm 4.x series properly. There is either something in the known problems or just unexpected bugs happen. WTF.

[andre30correia]

I cant understand why navi is not supported, amd need to improve their support, hire more devs or something

[Mathias]

Just for some positive feedback: Rocm-OpenCL works for me on my 5700. I'm on the latest fedora and I am simply using the official AMD repository. I think I had to add /opt/... manually to my (LD_LIBRARY)path for it to work. But thats it. I didn't manage to install it for a long time, but with 4.1 I managed to manually install it and yesterday I tried the repo again and it worked without a problem.

[sweetsuicide]

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

[david-nk]

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.