Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Error occurred while compiling with cuda 11 #310

Closed
PermissionDenied7335 opened this issue Aug 1, 2020 · 6 comments
Closed

Error occurred while compiling with cuda 11 #310

PermissionDenied7335 opened this issue Aug 1, 2020 · 6 comments

Comments

@PermissionDenied7335
Copy link

Building NVCC (Device) object CMakeFiles/xmrig-cuda.dir/src/nvidia/Release/xmrig-cuda_generated_cuda_core.cu.obj
nvcc fatal : Unsupported gpu architecture 'compute_30'
CMake Error at xmrig-cuda_generated_cuda_core.cu.obj.Release.cmake:220 (message):
Error generating
C:/Users/grade/xmrig-nvidia/build/CMakeFiles/xmrig-cuda.dir/src/nvidia/Release/xmrig-cuda_generated_cuda_core.cu.obj

@PermissionDenied7335
Copy link
Author

Disable compute_30 support may work, but I don't know how to.

@sOblivionsCall
Copy link

also seeing the same issue, other similar issues comment compute_30 in cmake.config but i don't see that anywhere for this.

The root of the issue atleast for me appears to be that i am running a too new version of cuda (11) that has depricated compute_30. Am going to try to downgrade to cuda 10 and see how that goes

@sOblivionsCall
Copy link

that said this project hasn't been touched in 14 months so i am not hopeful to receive any kind of support.

@xmrig
Copy link
Owner

xmrig commented Oct 12, 2020

This project replaced by unified xmrig miner and CUDA plugin this bug already fixed xmrig/xmrig-cuda@d483b2b but you can specify cmake option -DCUDA_ARCH= it not require change the code https://github.com/xmrig/xmrig-nvidia/blob/master/cmake/CUDA.cmake#L48
Thank you.

@StefanL
Copy link

StefanL commented Mar 7, 2022

What he is trying to say is that the default cuda architecture built into the cmake scripts includes sm_30, compute_30, which is not supported by the cuda runtime and tools (sdk) after cuda 10.2, hence the error that you see nvcc fatal : Unsupported gpu architecture 'compute_30'.

The easy way to avoid this error is to specify the GPU architecture you want to target with the following:
cmake .. -DCUDA_ARCH="40[;50;60]", where the multiple architectures in square brackets are optional.

Check the nvidia documentation to see what cuda version supports your video card, and what the compute capability of your card is and make sure they match.

For using older graphics cards with sm_30 or sm_20 specs you need to use older cuda sdks as well as older compilers (ie. gcc 7.5 and older). It gets complicated keeping all these different development environments straight.

@PermissionDenied7335
Copy link
Author

PermissionDenied7335 commented Mar 12, 2022

It seems that I have solved the problem on my laptop but I forgot to close it. I'll close it anyway.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants