Dear All,
I have recompiled the source code by the legend way again.
This time, I only kept the sm80flags and removed the sm86flags.
It's still work.
According to Ross's comment, thus it may be not necessary to include the sm86flag at current moment.
+Farid.
I have a small system, including 40078 atoms by NPT (ntt=3, gamma_ln=5) ensemble for 5 ns production run.
I used intel complier2017 + Nvidia cuda 11.1 for compile.
I extracted the data form the summary of the calculation as below:
Nvidia 2080ti + CUDA 10.2 221.96 ns/day
Nvidia 2080ti + CUDA 11.1 246.85 ns/day
Nvida 3080 + CUDA 11.1 270.57 ns/day
So, it may provide about 10% enhancement roughly.
For the GPU performance, it think you can visit the benchmark test.
http://ambermd.org/GPUPerformance.php
https://blog.exxactcorp.com/rtx3090-benchmarks-for-hpc-amber-a100-vs-2080ti-vs-rtx6000/
Regards,
Kai
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Jan 02 2021 - 19:00:02 PST
