[CP2K-user] [CP2K:14156] Re: GPU testruns fails with: CUDA driver version is insufficient for CUDA runtime version

[Alfio Lazzaro]

OK, sorry, I was a bit vague in my previous post...

So, the totalchain flag sets the SM architecture of the GPU in the arch 
files used to compile CP2K.
You can check the local.psmp files for the relevant flags. They are 
documented at

https://github.com/cp2k/cp2k/blob/master/INSTALL.md#2j-cuda-optional-improved-performance-on-gpu-systems

Maybe some of the libraries have the wrong path? Could you try a small CUDA 
helloworld example and get it to work?

Alfio


Il giorno martedì 3 novembre 2020 alle 23:14:20 UTC+1 sassy ha scritto:

> Hi Alfio,
>
> thanks for the feedback. I actually was searching the internet and the 
> list 
> here (my archive does go back a few years) without luck. 
>
> There was a reason why asked here as well: When you are using the 
> toolchain 
> you need to supply which type of GPU card you have, like the V100 for 
> example 
> in my case (see a recent posting of me). Thus, it could be that there is a 
> problem with that specific tuning. It was a shot into the dark and I 
> missed 
> the target. 
>
> In any case, thanks for your help!
>
> All the best from a cold London
>
> Jörg
>
> Am Montag, 2. November 2020, 09:16:26 GMT schrieb Alfio Lazzaro:
> > Hi Jörg,
> > This problem seems unrelated to CP2K...
> > Have you tried to google that message?
> > I found some entries:
> > 
> > 
> https://stackoverflow.com/questions/41409842/ubuntu-16-04-cuda-8-cuda-driver
> > -version-is-insufficient-for-cuda-runtime-vers
> > 
> https://stackoverflow.com/questions/52346957/cudagetdevice-failed-status-cu
> > da-driver-version-is-insufficient-for-cuda-run
> > 
> https://forums.developer.nvidia.com/t/error-cuda-driver-version-is-insuffic
> > ient-for-cuda-runtime-version/66006
> > 
> > Any many others.... It seems a problem with the driver installation, 
> people
> > recommend to uninstall/reinstall...
> > 
> > Alfio
> > 
> > Il giorno lunedì 2 novembre 2020 alle 00:23:32 UTC+1 sassy ha scritto:
> > > Dear all,
> > > 
> > > over the weekend I had access to a GPU test machine so I could compare 
> the
> > > performance between a RTX-6000 and a A100 GPU card.
> > > 
> > > I have installed the following CUDA drivers:
> > > $ nvidia-smi
> > > NVIDIA-SMI 450.80.02 Driver Version: 450.80.02 CUDA Version: 11.0
> > > 
> > > I am using GCC-9.3.0 with CUDA-11.0.2
> > > 
> > > I am using the toolchain with the following commands:
> > > 
> > > $ ./install_cp2k_toolchain.sh -j 12 --no-check-certificate 
> --libint-lmax=6
> > > --
> > > with-cmake=system --with-libxc --with-libint --with-fftw=system --with-
> > > openblas=system --with-libxsmm --with-parmetis --with-superlu 
> --with-quip
> > > --
> > > with-plumed --with-gsl --with-libvdwxc --with-spglib --with-hdf5
> > > --with-spfft
> > > --with-scalapack=system --gpu-ver=V100 --with-elpa=no --with-sirius=no
> > > 
> > > (I had problems with the ELPA and SIRIUS build, hence I disabled them).
> > > 
> > > Running the test jobs, *all* of them fail with:
> > > 
> > > "CUDA driver version is insufficient for CUDA runtime version"
> > > 
> > > The box has both cards installed and I am masking one with the
> > > CUDA_VISIBLE_DEVICES
> > > environment variable.
> > > 
> > > I am puzzled of where I am getting wrong. Yesterday I tried it with 
> both,
> > > the
> > > A100 and the RTX-6000 card using most of the above modules being 
> installed
> > > on
> > > the system via EasyBuild, today I only tested the RTX-6000 card.
> > > 
> > > Does anybody has a good idea? I don't have the box for much longer, as 
> I
> > > said,
> > > it is only a test environment I got access granted to.
> > > 
> > > Thanks for your help!
> > > 
> > > All the best from a mild London
> > > 
> > > Jörg
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20201104/dc24b0b9/attachment.htm>