[gmx-users] GPUs performance on Gromacs calculation

[Mark Abraham]

Hi,

On Thu, May 12, 2016 at 8:26 AM Ming Tang <m21.tang at qut.edu.au> wrote:

> Dear experts,
>
> I did SMD on 2 M2090 GPUs and 16 CPUs on one node. The simulation ran
> around 500 steps every 15 minutes.
> However, the SMD simulation ran around 68000 steps using 96 CPUs without
> GPU.
> Is this because the M2090 GPU type works badly on Gromacs?
>

We can't possibly tell what's going on - you've reported a rate and a total
number of steps for two different runs :-) If you want some feedback then
we need to know things like GROMACS version, what you are simulating, and
for you to share some .log files on a file-sharing service. The list does
not accept attachments


> Our Uni has the following types of GPU. Could anybody help to tell me
> which one is better for my SMD simulation?
>
> *         4 x M2050 Nvidia Tesla GPUs
> *         8 x M2090 Nvidia Tesla GPUs (2 per node)
> *         12 x K40 Nvidia Tesla GPUs (2 per node)
> *         8 x Intel Xeon Phi 5110P Co-processors (2 per node)
> *         1 x Bitware S5-PCIe-HQ with Altera Stratix V FPGA
>

The Teslas are probably all fine, the Phis probably useful only for
reaction-field simulations, and the FPGA useless.

Mark


> Thank you.
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>