Hello everyone,
My computer characteristics used to be:
Ryzen 1700
8 GB RAM
Nvidia GTX 960
Gromacs 2018.2 uses 100% of the CPU and 93-97% of the GPU for a protein system solvated in water and counter ions(65000 atoms) so it was good.
Recently a big donation with Titan XP arrived.
Running the same system on the GPU uses only 50-55%, yet uses 100% CPU.
Should I recompile Gromacs for the new GPU?
Sincerely,
Hovakim Grabski
Ph.D. student at Russian-Armenian University
GROMACS uses both CPU and GPU for different tasks, so just adding a more powerful GPU may not help if the workload is already limited by the CPU. Look at the end of the log file for suggestions. If your runs were not using PME on a GPU, then that is the main thing you can do that might help. But for more concrete suggestions, please attach the two log files for equivalent runs.
There’s more background information and tips at http://manual.gromacs.org/documentation/current/user-guide/mdrun-performance.html#running-mdrun-with-gpus.
Dear Mr. Abraham,
Thank You very much for Your suggestion.
Here are the log files using the same tpr runs.
for CMAKE I used:
cmake … -DGMX_GPU=ON -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-8.0 -DGMX_BUILD_OWN_FFTW=ON -DCMAKE_BUILD_TYPE=Release -DREGRESSIONTEST_DOWNLOAD=ON -DBUILD_SHARED_LIBS=OFF -DCMAKE_C_COMPILER=gcc -DGMX_MPI=OFF
Sincerely,
Hovakim Grabski