[Minor Feature Request] Multi-GPU status, API-specific Stress Test, and BIOS/CUDA details

[try228]

To bring GPU-T even closer to (and beyond) GPU-Z functionality on Linux, I'd like to suggest a few additions for future phases:

1. Multi-GPU & Hybrid Graphics Status:

• Add a status indicator for Multi-GPU setups: SLI/NVLink (NVIDIA), Crossfire(X) (AMD).
• Also, include a toggle or info for Hybrid Graphics (iGPU + dGPU), which is very common for Linux laptop users.

2. API-based Stress Test:

• Instead of a simple render test, implement a small GPU stress/load tool that can run on different APIs: Vulkan, OpenGL, OpenCL.
• Optionally, add detection for translation layers like Zink or Proton during the test to see how the GPU handles translated workloads.
• Note: I understand this is a complex task. If you're not sure how to implement the rendering part yet, or if it's too time-consuming, you could always keep it as a "community task" for later or ask for help from someone who specializes in Vulkan/OpenGL. No pressure!

3. Advanced Tab Additions:

• BIOS Details: Display more granular information about the video BIOS
• CUDA/ZLUDA Info: Since I'm using ROCm for ZLUDA, it would be great to see specific ZLUDA/CUDA versioning and status in the Advanced tab.

[lseurttyuu]

Moving into Discussions now, we'll discuss it there.

[lseurttyuu]

Thanks for the comment! Now, about these FRs:

1. Multi-GPU is essentially dead nowadays. Won't happen in GPU-T.

What do you mean by "include a toggle or info for Hybrid Graphics"? I'm not sure I understand the idea / the point of the idea.

2. This sounds like having a complex knowledge about graphical APIs is a must, which I do not have at the moment. However this idea is definitely an interesting one, I'll keep the discussion open for that.

3. BIOS details are nice to have, but pain to implement. There's probably a lot of low-level programming involved which I'm not sure I want to go for. Definitely not at the moment.

Regarding CUDA/ZLUDA: I'm going to include CUDA information for NVIDIA GPUs in the Advanced tab. I'll probably skip ZLUDA as it's not official / not tied directly with a GPU. It's a third-party translation layer between ROCm and CUDA which does not mean a certain set of GPU features/capabilities, doesn't have utilities like clinfo (as far as I know).

[try228]

What do you mean by "include a toggle or info for Hybrid Graphics"? I'm not sure I understand the idea / the point of the idea.

What i meant is features like NVIDIA Optimus or analogs that can use both dGPU and iGPU(specifically on laptops, but does appear on desktop PCs too)

Regarding CUDA/ZLUDA: I'm going to include CUDA information for NVIDIA GPUs in the Advanced tab. I'll probably skip ZLUDA as it's not official / not tied directly with a GPU. It's a third-party translation layer between ROCm and CUDA which does not mean a certain set of GPU features/capabilities, doesn't have utilities like clinfo (as far as I know).

Alright, as long as ZLUDA is not included in the main ROCm stack you can not add it, but add atleast info for ROCm(AMD) and oneAPI Level Zero(Intel) same as for CUDA, as those are presented as competing technologies similiar to CUDA.

[lseurttyuu]

What i meant is features like NVIDIA Optimus or analogs that can use both dGPU and iGPU(specifically on laptops, but does appear on desktop PCs too)

To sum it up - I believe @Danny3 described is nicely in #33 . Integrated/Discrete info should be enough and definitely nice to have.

Alright, as long as ZLUDA is not included in the main ROCm stack you can not add it, but add atleast info for ROCm(AMD) and oneAPI Level Zero(Intel) same as for CUDA, as those are presented as competing technologies similiar to CUDA.

ROCm info is something definitely worth considering. AMD provides rocminfo tool which displays various information probably worth parsing & displaying in GPU-T. I don't know anything about Intel's oneAPI Level Zero (yet), but that will be for a later time.

[try228]

ROCm info is something definitely worth considering. AMD provides rocminfo tool which displays various information probably worth parsing & displaying in GPU-T. I don't know anything about Intel's oneAPI Level Zero (yet), but that will be for a later time.

I could actually help you with oneAPI, as my iGPU supports it(but though, not matrix acceleration(as it has no XMX engines like Arc does), only vector.)