[gmx-users] Tests with Threadripper and dual gpu setup

[Harry Mark Greenblatt]

BS”D



Did you build with or without hwloc?

I did use hwloc.




—————————————————
Gromacs 2018 rc1 (using gcc 4.8.5)
—————————————————

Using AVX_256


You should be using AVX2_128 or AVX2_256 or Zen! The former will be fastest
in CPU-only runs, the latter can often be (a bit) faster in GPU accelerated
runs.


Once I saw that AVX2_128 was faster, I did not think there would be an advantage to AVX2_256 on GPU runs.

So is there any suggestion to overcome the problem of gcc 5.5 not recognising what CPU hardware I have (not that 5.5 gave much of an advantage in Gromacs 2016).


Now force Dynamic Load Balancing

gmx mdrun -v -deffnm test.npt -s test.npt.tpr -pme gpu -pin on -ntmpi 4
-npme 1 -gputasks 0011 -nb gpu -dlb yes


I would recommend *against* doing that unless you have concrete cases where
this is better than "-dlb auto" -- and if you have such cases, please share
them as it is not expected behavior. (Note: DLB has acquired the capability
to observe when turning it on it leads to performance drop and it switches
off automatically in such cases!)


I did see that it in some cases it was turning off DLB for a while, or for the rest of the run.

In my case however, I did get better results by forcing it to be on.

I can send the .tpr file to you, off-list, if you want…



Thanks

Harry




--------------------------------------------------------------------
Harry M. Greenblatt
Associate Staff Scientist
Dept of Structural Biology           harry.greenblatt at weizmann.ac.il<../../owa/redir.aspx?C=QQgUExlE8Ueu2zs5OGxuL5gubHf97c8IyXxHOfOIqyzCgIQtXppXx1YBYaN5yrHbaDn2xAb8moU.&URL=mailto%3aharry.greenblatt%40weizmann.ac.il>
Weizmann Institute of Science        Phone:  972-8-934-6340
234 Herzl St.                        Facsimile:   972-8-934-3361
Rehovot, 7610001
Israel