[gmx-users] cuda problem

[Szilárd Páll]

On Tue, Jul 9, 2013 at 11:20 AM, Albert <mailmd2011 at gmail.com> wrote:
> On 07/09/2013 11:15 AM, Szilárd Páll wrote:
>>
>> Tesla C1060 is not compatible - which should be shown in the log and
>> standard output.
>>
>> Cheers,
>> --
>> Szilárd
>
>
> THX for kind comments.
>
> do you mean C1060 is not compatible with cuda-5.0 toolkit? or it is not
> compatable with Gromacs-4.6.3?I only obtained the following information in
> the log file
>
>
> 2 GPUs detected on host c0107:
>   #0: NVIDIA Tesla C2075, compute cap.: 2.0, ECC: yes, stat: compatible
>   #1: NVIDIA Tesla C2075, compute cap.: 2.0, ECC: yes, stat: compatible
>
> 2 GPUs auto-selected for this run: #0, #1

Those are not C1060-s (which is compute capability 1.3, hence not
compatible with GROMACS), but C2075-s!

>
> NOTE: Using a GPU with ECC enabled and CUDA driver API version <5.0, known
> to
>       cause performance loss. Switching to the alternative polling GPU wait.
>       If you encounter issues, switch back to standard GPU waiting by
> setting
>       the GMX_CUDA_STREAMSYNC environment variable.

As I said before, this refer to the NVIDIA driver. You can ignore
this, but you'd be better off with newer drivers.

>
> Non-default thread affinity set probably by the OpenMP library,
> disabling internal thread affinity

This means that either you, or the job scheduler set GOMP_CPU_
AFFINITY or KMP_AFFINITY. This may affect performance negatively, but
you'll have to try yourself.

--
Szilárd

>
>
> best
>
> Albert
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the www
> interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists