Radeon ROCm 4.1 Released - Still Without RDNA GPU Support

[qarium]

Nice if they can give us a "prosumer" CDNA card that doesn't cost 5 grand but still gives us decent compute.

as soon as they have chiplet design they can easily do this.
imagine a full version has 4 chips so they cut out 3 chips and you get a card for like the 1/4 of the price.

best case szenario for consumers would be a RDNA/CDNA chiplet design hybrid...
it is not technically impossible to build this.

some people claim it makes no sense to split GCN into RDNA and CDNA if you put these 2 chips together in a chiplet design hybried of RDNA+CDNA...

but i think they are wrong because General Compute if you care about performance is death and more specialized chips will rule the performance.

[vegabook]

https://www.amd.com/system/files/doc...whitepaper.pdf

Nice if they can give us a "prosumer" CDNA card that doesn't cost 5 grand but still gives us decent compute.

[bridgman]

I was wondering how a CDNA would work with DRI_PRIME=1 so I guess that answers that. I guess I won't be buying those used to supplement desktop APUs

Yeah, they would not be good for supplementing APU graphics, although they would be really good for supplementing APU compute.

[skeevy420]

Mind if I tweak this a bit ? I think it's important to distinguish between "no display connector" (with or without display HW) and "no graphics pipeline".

We do sell a lot of headless graphics cards for remote gaming and virtual desktop applications, while the compute cards are unable to do graphics even remotely.

I was wondering how a CDNA would work with DRI_PRIME=1 so I guess that answers that. I guess I won't be buying those used to supplement desktop APUs

[bridgman]

Both are an evolution of GCN and the biggest difference between the two is one is designed for graphics output and the other is designed for computing. Laymanized -- headless and not headless.

Mind if I tweak this a bit ? I think it's important to distinguish between "no display connector" (with or without display HW) and "no graphics pipeline".

We do sell a lot of headless graphics cards for remote gaming and virtual desktop applications, while the compute cards are unable to do graphics even remotely.

[skeevy420]

https://www.amd.com/system/files/doc...whitepaper.pdf

From the same paper:

Unlike the graphics-oriented AMD RDNA™ family, the AMD CDNA family removes all of the fixed-function hardware that is designed to accelerate graphics tasks such as rasterization, tessellation, graphics caches, blending, and even the display engine

Or Headless RDNA.

From the RDNA whitepaper:

The new RDNA architecture is optimized for efficiency and programmability while offering backwards compatibility with the GCN architecture. It still uses the same seven basic instruction types: scalar compute, scalar memory, vector compute, vector memory, branches, export, and messages. However, the new architecture fundamentally reorganizes the data flow within the processor, boosting performance and improving efficiency.

Both are an evolution of GCN and the biggest difference between the two is one is designed for graphics output and the other is designed for computing. Laymanized -- headless and not headless.

[pete910]

Yep... the message is a bit misleading though... what it is supposed to be saying is "you can't run 4.1 userspace without the 4.1 kernel driver".

The required kernel driver changes are on their way upstream but not there yet AFAIK, so just going to newer kernel isn't going to help. I was going to say "if you can go back to 4.0 userspace in the short term that would probably be best" but you already said that you had problems with 4.0. Do you happen to remember what problems you saw ?

The problem was rocminfo listed and stated both GPU's and cpu fine. But when something like f@h, luxmark ect tried to use any GPU it failed on both GPU's
CPU worked fine.

[raun0]

For all intents and purposes, CDNA is basically headless RDNA; both are the successor to GCN.

https://www.amd.com/system/files/doc...whitepaper.pdf

The classic GCN compute cores contain a variety of pipelines optimized for scalar and vector instructions. In particular, each CU contains
a scalar register file, a scalar execution unit, and a scalar data cache to handle instructions that are shared across the wavefront, such as
common control logic or address calculations. Similarly, the CUs also contain four large vector register files, four vector execution units that
are optimized for FP32, and a vector data cache. Generally, the vector pipelines are 16-wide and each 64-wide wavefront is executed over
four cycles.
The AMD CDNA architecture builds on GCN’s foundation of scalars and vectors and adds matrices as a first class citizen while
simultaneously adding support for new numerical formats for machine learning and preserving backwards compatibility for any software written for the GCN architecture. These Matrix Core Engines add a new family of wavefront-level instructions, the Matrix Fused Multiply-
Add or MFMA. The MFMA family performs mixed-precision arithmetic and operates on KxN matrices using four different types of input
data: 8-bit integers (INT8), 16-bit half-precision FP (FP16), 16-bit brain FP (bf16), and 32-bit single-precision (FP32). All MFMA instructions
produce either 32-bit integer (INT32) or FP32 output, which reduces the likelihood of overflowing during the final accumulation stages of a
matrix multiplication.

[bridgman]

Looks to be that issue as according to rocminfo it's disabling it and I need to upgrade AMDGPU

Yep... the message is a bit misleading though... what it is supposed to be saying is "you can't run 4.1 userspace without the 4.1 kernel driver".

The required kernel driver changes are on their way upstream but not there yet AFAIK, so just going to newer kernel isn't going to help. I was going to say "if you can go back to 4.0 userspace in the short term that would probably be best" but you already said that you had problems with 4.0. Do you happen to remember what problems you saw ?

[smitty3268]

Seem to recall reading that has now terminated last year IIRC but maybe completely wrong.

The last I saw they have an agreement with specific targets and prices through the end of this year. Then there are additional agreements through until March 1st 2024, which seem a little vaguer and are possibly still up for some negotiation.