Hacker News new | past | comments | ask | show | jobs | submit login

> they choose it because it works

MS DirectCompute also works. Yet last time I checked, MS Azure didn’t support DirectCompute with their fast GPUs. These virtual machines come with TCC (Tesla Compute Cluster) driver which only supports CUDA, DirectCompute requires a WDM (Windows Driver Model) driver. https://social.msdn.microsoft.com/forums/en-US/2c1784a3-5e09...




You can flip the model from TCC to WDDM via nvidia-smi.

But AFAIK, C++ AMP is deprecated and going away.

https://docs.microsoft.com/en-us/cpp/parallel/amp/cpp-amp-ov...

> C++ AMP headers are deprecated, starting with Visual Studio 2022 version 17.0. Including any AMP headers will generate build errors. Define _SILENCE_AMP_DEPRECATION_WARNINGS before including any AMP headers to silence the warnings.

So please don't rely on DirectCompute. It's firmly in legacy territory. Microsoft didn't invest the effort necessary to make it thrive.


> flip the model from TCC to WDDM via nvidia-smi

I’m not sure that’s legal. I think NV wants extra money, details there https://www.nvidia.com/content/dam/en-zz/Solutions/design-vi...

> C++ AMP is deprecated and going away.

Indeed, but I never used that thing.

> don't rely on DirectCompute

DirectCompute is a low-level tech, a subset of D3D11 and 12. It’s not deprecated, used by lots of software, most notably videogames. For instance, in UE5 they’re even rasterizing triangles with compute shaders, that’s DirectCompute technology.

Some things are worse than CUDA. Different programming language HLSL, manually managed GPU buffers, compatibility issues related to FP64 math support.

Some things are better than CUDA. No need to install huge third-party libraries, integrated with other GPU-related things (D2D, DirectWrite, desktop duplication, media foundation). And vendor agnostic, works on AMD and Intel too.


> I’m not sure that’s legal. I think NV wants extra money, details there https://www.nvidia.com/content/dam/en-zz/Solutions/design-vi...

Use: nvidia-smi -g {GPU_ID} -dm 0

Cloud providers do pay for an extensive vGPU license, don't worry about that part.

> DirectCompute is a low-level tech, a subset of D3D11 and 12.

D3D11. The compute subset of D3D12 is named D3D12 and directly got rolled into that. Also, you have CLon12 today which does support SPIR-V.


> Use: nvidia-smi -g {GPU_ID} -dm 0

I think I tried that a year ago, and it didn’t work. Documentation agrees, it says “GRID drivers redistributed by Azure do not work on non-NV series VMs like NCv2, NCv3” https://docs.microsoft.com/en-us/azure/virtual-machines/wind... Microsoft support told me the same. I wanted NCv3 because on paper, V100 GPU is good at FP64 arithmetic which we use a lot in our compute shaders.

> The compute subset of D3D12 is named D3D12

Interesting, didn’t know about the rebranding.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: