AMD Has Open-Source Ryzen AI Demo Code - But Only For Windows

[rene]

lol, if just somebody told them that nobody is using Win~DOS anymore ;-)!

[boboviz]

Simple, they observed the blind devotion that Linux users have for nvidia even when they are hostile towards FOSS, so they are assuming that we respond better when ignored/abused.

That's simply....TRUE!

[Linuxxx]

lol, if just somebody told them that nobody is using Win~DOS anymore ;-)!

In reality though, nobody (except you) is using that Berlin-hipster T2 SDE of yours!

In actually relevant news:

SteamOS 3.5 will most likely premier in preview form this week...

[boboviz]

I suspect AMD actual Linux capable personnel is very small and mostly focused on AMDGPU kernel coding which would help explain RoCm horror show

Rocm is great for HPC (es. Frontier and LUMI), but seems they are not interested in the "workstation/pc ecosystem"

[rene]

In reality though, nobody (except you) is using that Berlin-hipster T2 SDE of yours!

In actually relevant news:

SteamOS 3.5 will most likely premier in preview form this week...

How is my OS related to AMDs Windows software? In more relevant news, being 25 years in the making with developers from all around the world T2 can barely be called a Berlin-hipster OS. Heck even I often developer from other parts of the world! And as long as it is the BEST OS for myself and we have more users than Serpent and Temple OS combined that's all I need ;-) ¯\_(ツ)_/¯

[marlock]

question 1:
is this extra IP a gimmick or can it really do things that are way less efficient in general purpose cpu cores and gpus? isn't it the sort of task an avx512 matrix processing could do well?

question 2:
does it look like this is meant to be packed into the cpu or the gpu in desktops (similar to video encode/decode blocks)?

[GreenReaper]

I suspect it can't do it as efficiently, hence debuting on mobile (plus they can get an AI laptop for a developer). But yeah, would be interested to know how it compares to, say, six Zen 4 cores - or sixteen.

[Spacefish]

It runs perfectly fine on linux when changing a few paths from \ to /..
I tried it and it works . Even did a PR to add "search on enter" feature, but no one replied to it. So i never made the effort to make it cross-platform and do a PR as it won´t be merged probably..

The accelerator is not used of course (no hardware there yet anyway, 7040 series laptop availiable).. But as it´s ONNX based, it will run just fine on your GPU or CPU or any other Onnx supported Accelerator.

But it should be trivial for AMD to support it on linux, as it seems to be essentially based on the Xilinx Versal/Vitis AI Core supported in accelerator cards since years in linux: https://www.xilinx.com/products/tech...ai-engine.html

The ONNX runtime component for Linux already exists for other Xilinx Platforms: https://onnxruntime.ai/docs/executio...nProvider.html shouldn´t be that hard for AMD to enable the Phoenix integrated accelerator on Linux.

Furthermore the comment "Ryzen AI Linux support is not enabled in this release" implies that there already is support but it´s just not ready for primetime / not enabled by default.. Probably like RDNA is not officially supported on ROCm but still works somehow..

[qarium]

Rocm is great for HPC (es. Frontier and LUMI), but seems they are not interested in the "workstation/pc ecosystem"

did you not work with the community on compute the last 10 years ?

Not to the same extent that we did for graphics. Our open source graphics focus was desktop users on common distros, while the compute focus was on large corporate customers who did not use standard distros or work in the community to the same extent.

i did buy amd hardware for gpu compute in the last 10 years... that AMD did only care for "large corporate customers who did not use standard distros" ... and litterally ignored the community does in my point of view hit them hard now with all the AI hype and all the people buy Nvidia cards for AI workloads on CUDA...

AMD has to end this negative elitism focus on: "large corporate customers" because it damange the reputation of the AMD brand name.

i personally want to buy AMD hardware for AI workloads without AMD telling me they are only for "large corporate customers"
thats nonsense.

AMD should think about this; why not ship opensource AI modells inside the driver and make these AI functions easily accessible to other software like games. there are some very interesting opensource AI modells out there.

AMD should think about dualgpu cards again with one chip RDNA3 and the other chip CDNA
also producing a smaller CDNA die but with very large amount of ram to avoid the memory wall in AI tools could be a top seller as a second card to the existing graphic card.

i also would find it nice to see some FPGAs inclused in any CPU/APU the Mega65(comodore64) and VampirV4(Amiga) proof that there is a market for FPGAs to emulate old computer systems for games. and honestly a virtual machine without FPGA does not work well for that. the FPGA should be large enough to emulate 8bit/16bit and maybe some popular 32bit area gaming platforms.

[bridgman]

But why Windows when the majority of such loads happen on Linux........................

I don't agree - certainly network training is majority Linux but network usage / inference happens much closer to end users and so Windows plays a larger role.