Dear All,
Thanks to a close collaboration with NVIDIA I am pleased to announce that we
now have support for GPU acceleration (using NVIDIA CUDA) in PMEMD. At
present this only covers GB simulations but we hope to add PME support
shortly.
Depending on system size using a GPU can be between 10 and 80x faster than a
single 2.8GHz Intel processor.
I have put together a webpage that shows provisional benchmarks along with
patches for AMBER 10. Support and test cases are already included in the
AMBER 11 CVS Tree.
http://ambermd.org/gpus/index.htm
As of Sept 4th 2009 I have tested this code on the following NVIDIA GPUS:
Tesla C1060
GTX295
Additionally it should work on any v1.3 NVIDIA GPU including:
Tesla C1070
GTX285
GTX275
At present only a single GPU is supported but it is our intention to add
support for multiple GPUs (e.g. C1070) shortly.
Scott Le Grand at NVIDIA has worked tirelessly to ensure that results match
as closely as possible with the CPU results. Testing shows that energy
conservation is exactly equivalent to the CPU code. However, extensive
testing is still required.
Compilation is easy. Assuming you have the NVIDIA CUDA drivers and CUDA
toolset installed you just configure PMEMD with:
./configure linux_em64t_cuda gfortran nopar pubfft
Please feel free to try the patch and/or the AMBER 11 tree and let me know
your feedback.
All the best
Ross
/\
\/
|\oss Walker
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk | PGP Key available on request |
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Sep 14 2009 - 11:51:05 PDT
