Slow MD simulation and problem with GPU support

GROMACS version: 2021.1
My PC has features of GPU NVIDIA geoforce GTX 1660 Ti, and CPU Cori 7 (10 GEN) 10870, 16 logical cores, base speed 2.2 GHz up to 5 GHZ. My MD simulation is very slow around 4 ns/day is this consider normal for such PC ?!. and when i try to run GMX on GPU it said as the following picture (;

, to be informed i did all the following commands gcc 7.5.0, g++, -DGMX_MPI=on, -DGMX_GPU=CUDA, -DGMX_BUILD_OWN_FFTW=ON -DREGRESSIONTEST_DOWNLOAD=ON and it is said the whole things are compiled. when i tried to run gmx mdrun -deffnm md10 -nb gpu, it said “but the GROMACS
binary has been built without GPU support”. please help

Hi,

it’s impossible to gauge whether the performance is good or bad without knowing more about your system. 4ns/day doesn’t seem outrageously bad for a single workstation, depending on your simulated system of course – but your message does suggest that a few performance pitfalls are there.

Are you running the Ubuntu provided Gromacs, or the one you compiled yourself? The log suggests to me that the used binary is pre-compiled by Ubuntu (it has the distribution name in the “version” field at the top of the log), which will not use your GPU and use the slowest common denominator for perfomance optimizations (i.e. SSE2). To check this, uninstall any version installed through the Ubuntu package manager, and be sure to source GMXRC from the Gromacs installation directory.

Regards,
Petter

Many thanks, the problem was in the source I haven’t use it properly