Hi All,
Here's some initial RTX5080 benchmarks for Amber 24. This is with CUDA 12.8 and Driver v570.86.16.
JAC_PRODUCTION_NVE - 23,558 atoms PME 4fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 1513.55 seconds/ns = 57.08
JAC_PRODUCTION_NPT - 23,558 atoms PME 4fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 1468.05 seconds/ns = 58.85
JAC_PRODUCTION_NVE - 23,558 atoms PME 2fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 815.23 seconds/ns = 105.98
JAC_PRODUCTION_NPT - 23,558 atoms PME 2fs
-----------------------------------------
[0] 1 x GPU: | ns/day = 764.65 seconds/ns = 112.99
FACTOR_IX_PRODUCTION_NVE - 90,906 atoms PME
-------------------------------------------
[0] 1 x GPU: | ns/day = 394.81 seconds/ns = 218.84
FACTOR_IX_PRODUCTION_NPT - 90,906 atoms PME
-------------------------------------------
[0] 1 x GPU: | ns/day = 365.36 seconds/ns = 236.48
CELLULOSE_PRODUCTION_NVE - 408,609 atoms PME
--------------------------------------------
[0] 1 x GPU: | ns/day = 105.96 seconds/ns = 815.43
CELLULOSE_PRODUCTION_NPT - 408,609 atoms PME
--------------------------------------------
[0] 1 x GPU: | ns/day = 99.07 seconds/ns = 872.08
STMV_PRODUCTION_NPT - 1,067,095 atoms PME
-----------------------------------------
[0] 1 x GPU: | ns/day = 63.17 seconds/ns = 1367.76
TRPCAGE_PRODUCTION - 304 atoms GB
---------------------------------
[0] 1 x GPU: | ns/day = 1471.73 seconds/ns = 58.71
MYOGLOBIN_PRODUCTION - 2,492 atoms GB
-------------------------------------
[0] 1 x GPU: | ns/day = 871.89 seconds/ns = 99.09
NUCLEOSOME_PRODUCTION - 25,095 atoms GB
---------------------------------------
[0] 1 x GPU: | ns/day = 30.20 seconds/ns = 2860.68
In summary not great... Indeed pretty terrible to be honest. Actually ends up about the same / a little slower than the RTX4080.
https://www.exxactcorp.com/blog/molecular-dynamics/amber-molecular-dynamics-nvidia-gpu-benchmarks
This may improve with newer drivers and or later CUDA versions but in short this is the first new GPU model out of the box that wasn't automatically ~30% faster than the predecessor - which doesn't bode well for the future unfortunately. :-(
All the best
Ross
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Feb 12 2025 - 13:30:02 PST
