I'm doing equilibrium and steered molecular dynamics simulations with GROMACS 4.6.5 in a single computer with 2 GPUs and 8 CPUs. I have obtained that simulations in the GPU-CPU goes almost 3 times faster than when they are run only at CPU. I would like to know whether this performance is reasonable or may it be increased?.
The last section of the .log file will have performance statistics so you can see what was taking the most time. The balance between the speed of the GPU and the CPUs is critical to achieving good performance.
Thank's for the answer. I have checked the .log file and there is a line with the "force evaluation time GPU/CPU=6.882 ms/7.277 ms = 0.946" and also says that for optimal performance this ratio should be close to one. This indicate I'm obtaining the maximum performance? I though I would be able to obtain higher pèrformance when using GPU (higher than the 3x I'm obtaining now).
Yes, in that case that seems to be the optimal performance for your hardware.
For GPU acceleration your system size (# of particles) is crucial. The performance increase going from CPU to GPU is bigger for bigger systems as it profits from decompositions with less communication between your processes on the GPU. For large systems your will have more than a threefold boost. How many particles do you have?
Thank's for the answer. My system is around 70 000 atoms: 700 atoms are from the DNA and the rest are for the solvent.
I have a similar issue where my force evaluation time is around 5. what can i do to increase my performance
People looking to understand the nuts and bolts of GROMACS performance and how to get the most out of GPUs should (1) use the most recent software, rather than 4.6.5 which is quite outdated at this point and (2) read the following papers:
http://dx.doi.org/10.1002/jcc.24030
http://dx.doi.org/10.1007/978-3-319-15976-8_1
http://pubman.mpdl.mpg.de/pubman/item/escidoc:2037317/component/escidoc:2037318/2037317.pdf?mode=download
The performance depends on the quality of the hardware and the balance between CPU and GPU workloads. One cannot necessarily slap on an arbitrary GPU and expect massive gains.
- Badges
- Science topic
- Similar topics
- Chemistry
- Theoretical Chemistry
- Computational Chemistry
- Molecular Dynamics