Re: [AMBER] fail when run pmem.cuda

From: Carlos <carlos.rom.74he.gmail.com>
Date: Sun, 29 Nov 2020 23:56:43 -0600

Hi. Thanks for your support.

Yes, I ran the test suite and 0 tasks passed. File.out shows that GPU
was not detected. So I think you have reason in that the problem is the
old amber14 version.
I will try to get an actualization.

Thanks again.

Regards

El dom., 29 de noviembre de 2020 12:35 p. m., David A Case <
david.case.rutgers.edu> escribió:

> On Sat, Nov 28, 2020, Carlos wrote:
> >
> >I used to work with amber14 GPU code, ubuntu14, cuda6.5 GTX 970 and
> usually
> >it works well,, but sometimes my computer fails, so I bought a new CPU
> >and a GTX 1660 and I installed amber14 GPU code, ubuntu14.04, cuda 7.5
> and
> >installation was ok, but when I run a molecular dynamic, appears the next
> >message:
> >
> >cudaMemcpyToSymbol: SetSim copy to cSim failed invalid device symbol
>
> Did you run the test suite? That will help decide whether the problem
> is with the installation, or with your particular input(s). But I'm
> guessing that cuda7.5 may not produce valid code for Turing GPU's --
> that claim has been made on the mailing list, e.g. here:
>
> http://archive.ambermd.org/201904/0045.html
>
> But it is certainly possible that Amber14 won't run correctly on a GPU
> that was made more than 5 years later, even if one worked on getting
> around the fact that it thinks more recent CUDA versions are unsupported.
> You should certainly consider an upgrade to Amber20.
>
> ....dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Nov 29 2020 - 22:00:01 PST
Custom Search