[gmx-users] pp/pme ratio differences between gromacs 5.0.x and 5.1.x
johannes.wagner at h-its.org
Tue Feb 9 12:04:07 CET 2016
yes, that reproduces the performance. Of course I can always manually fine-tune pp/pme ratio. It’s just that gromacs 5.0 gave a quite good out of the box pp/pme ratio that allowed to submit the same job with [x] more nodes without thinking about it much. Now it seems we always needs to test for each change in run parameters. Be it box size, number of nodes, different queue with different core configuration. And thats what my question was, if it is intended behavior that gromacs 5.1 does estimate the pp/pme ratio in a different way than gromacs 5.0 for an identical system.
> you could try to manually set the number of PME nodes with -npme in 5.1.
> Does that reproduce the 5.0 performance?
>> On 07 Feb 2016, at 19:57, Johannes Wagner <johannes.wagner at h-its.org <mailto:johannes.wagner at h-its.org>> wrote:
>> hey guys,
>> came across an issue with pp/pme ratio difference from 5.0 to 5.1. For a system running on 2 nodes with a pp/pme ratio of 80:16 on 5.0, the same thing starts with 72:24 on gromacs 5.1. The 5.1 run has a much higher load imbalance and is hence slower. Are there known changes from 5.0 to 5.1 in the pp/pme ratio that cause this?
>> cheers, johannes
More information about the gromacs.org_gmx-users