From: Gawtak Kim (shootgoaltag_at_gmail.com)
Date: Thu Jul 29 2021 - 02:23:09 CDT
Hi,
I have experience running molecular dynamics simulations using
“NAMD_2.14_Win64-multicore”.
At that time, I ran NAMD by typing "namd2 +auto-provision" into the
Command Prompt, and as a result,
CPU 0~39 (40 logical processors) of NUMA Node 0 were used for computation,
while CPU 0~39 (40 logical processors) of NUMA Node 1 were not used
for computation.
For your information, the specifications of the workstation I used at
that time are as follows.
OS: Windows 10 Pro
CPU: Intel Xeon Gold 6242R 3.10 GHz (2 CPUs)
20 physical cores per CPU
2 logical processors per physical core
2 CPUs * 20 cores * 2 logical processors = 80 logical processors
GPU: NVIDIA Quadro P2200 (1 GPU)
RAM: 512GB
However, I want to use both CPU 0~39 (40 logical processors) of NUMA
Node 0 and CPU 0~39 (40 logical processors) of NUMA Node 1 at the same
time
to use the total of 80 logical processors simultaneously.
What should I enter in Command Prompt to make this possible?
If this is possible, will the simulation speed be twice as fast as the
simulation speed at that time?
Also, as far as I know, the NVIDIA Quadro P2200 GPU supports CUDA GPU
Acceleration.
If I use “NAMD_2.14_Win64-multicore-CUDA”, can I expect an increase in
simulation speed due to CUDA GPU Acceleration?
What should I enter in Command Prompt to make this possible?
If I should use both 80 logical processors and CUDA GPU Acceleration
to maximize simulation speed, what should I enter in Command Prompt to
make this possible?
Do I need to install one more identical GPU on the workstation to
maximize simulation speed? In other words, should the number of CPUs
and GPUs be the same?
I'd appreciate it if you could answer my question.
Thank you.
Best regards,
Gawtak Kim
This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST