Dear All
I followed the instructions, recompiled with the new patch and completed
the run that was crashing previoulsy (see emails to the list with the
subject "pmemd24 problems with RTX5090" ) .
I compared pressure, etot, temperature and density with exactly the same
run with a 4090 and results are quite similar. (si below for Final
Performance Info)
Just one issue. At some time the following message appeared in the
stdout / stderr:
"Note: The following floating-point exceptions are signalling: IEEE_UNDERFLOW_FLAG IEEE_DENORMAL"
I'm guessing this is not important, but still, It will be helpful if
anyone can confirm it
For 5090:
Final Performance Info:
-----------------------------------------------------
Average timings for last 1 steps:
Elapsed(s) = 0.00 Per Step(ms) = 0.08
ns/day = 1082.25 seconds/ns = 79.83
Average timings for all steps:
Elapsed(s) = 20557.91 Per Step(ms) = 2.06
ns/day = 42.03 seconds/ns = 2055.79
-----------------------------------------------------
Setup CPU time: 3.30 seconds
NonSetup CPU time: 20477.11 seconds
Total CPU time: 20480.40 seconds 5.69 hours
Setup wall time: 3 seconds
NonSetup wall time: 20558 seconds
Total wall time: 20561 seconds 5.71 hours
4090:
Final Performance Info:
-----------------------------------------------------
Average timings for last 1 steps:
Elapsed(s) = 0.00 Per Step(ms) = 0.19
ns/day = 447.40 seconds/ns = 193.12
Average timings for all steps:
Elapsed(s) = 25950.41 Per Step(ms) = 2.60
ns/day = 33.29 seconds/ns = 2595.04
-----------------------------------------------------
Setup CPU time: 6.25 seconds
NonSetup CPU time: 25885.31 seconds
Total CPU time: 25891.56 seconds 7.19 hours
Setup wall time: 6 seconds
NonSetup wall time: 25951 seconds
Total wall time: 25957 seconds 7.21 hours
thank you very much
OCS
Oscar Conchillo Solé
Computational Biology Group
Data Center Manager, Sysadmin and Bioinformatics
Institut de Biotecnologia i Biomedicina (UAB)
Department of Genetics and Microbiology (UAB)
mail:Oscar.Conchillo.uab.cat
telf: 0034 93581 4431
On 5/9/25 00:16, David A Case via AMBER wrote:
> A note: if you plan to run pmemd.cuda on NVIDIA Blackwell GPUs (B200,
> RTX5090...) you will need to run the *update_pmemd* script (in your
> pmemd24_src folder) to get updates 1 and 3.
>
> See https://ambermd.org/AmberPatches.php for more information on this.
>
> These updates are not needed (but should not hurt) if you are using
> earlier
> NVIDIA GPUs.
>
> Thanks to David Clark, Patricio Barletta and Adrian Roitberg for help
> with
> creating these updates.
>
> ....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 Fri Sep 12 2025 - 08:00:03 PDT
