Re: Slow performance over multi-core processor and CUDA build

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Tue Jul 28 2020 - 08:04:52 CDT

Not sure why hyperthreading is mentioned, which is not supported by the
processor in question:
https://ark.intel.com/content/www/us/en/ark/products/186604/intel-core-i7-9700k-processor-12m-cache-up-to-4-90-ghz.html

Roshan, what are the system size and simulation parameters? It is possible
that the system is not suitable for a CPU-GPU hybrid scheme (possibly made
worse by using too many CPU cores). The Gromacs benchmark (which was
probably run in single precision and on the CPU) seems to suggest a rather
small system. Have you tried running a non-GPU build? Or the
GPU-optimized 3.0 alpha build?

For typical biological systems (of the order of 100,000 atoms) and running
over CPUs, Gromacs would be faster over a few nodes but scale over multiple
nodes less well than NAMD. The tipping point depends on the system and to
a lesser extent on the hardware makeup. I suggest you benchmark your
system thoroughly with both codes, and then decide.

Giacomo

On Tue, Jul 28, 2020 at 8:37 AM Norman Geist <norman.geist_at_uni-greifswald.de>
wrote:

> I’d say don’t use hyperthreading in HPC in general, nothing special about
> GPUs. You can assign your tasks/threads to physical core only, e.g. using
> taskset, but determining the correct mapping can be tricky.
>
>
>
> Bests
>
> Norman Geist
>
>
>
> *Von:* owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] *Im
> Auftrag von *Ashkan Shekaari
> *Gesendet:* Dienstag, 28. Juli 2020 14:25
> *An:* namd-l <namd-l_at_ks.uiuc.edu>; Roshan Shrestha <roshanpra_at_gmail.com>
> *Betreff:* Re: namd-l: Slow performance over multi-core processor and
> CUDA build
>
>
>
> *Dear Roshan,*
>
>
>
> *DO NOT USE** GPU along with cpu- hyperthreading *
>
>
>
>
> *--*
> *All the best,*
> *Ashkan Shekaari,*
> *Ph.D Candidate in Solid State Physics,*
> *K. N. Toosi University of Technology,*
> *Tehran, Iran*
>
>
> ------------------------------
>
> *From: *"Roshan Shrestha" <roshanpra_at_gmail.com>
> *To: *"namd-l" <namd-l_at_ks.uiuc.edu>
> *Sent: *Tuesday, July 28, 2020 2:31:14 PM
> *Subject: *namd-l: Slow performance over multi-core processor and CUDA
> build
>
>
>
> Hi everyone,
>
> I have a system with i7 9700k 8 core processor and rtx
> 2070 Nvidia GPU. My namd is cuda build; but I am extremely disappointed
> with its performance as it gives me like 0.21 ns/ day while with the same
> system gromacs give me the performance of more than 300 ns/day. I really
> can't fathom what's the issue here as namd2 is using the GPU when I see the
> GPU processes using nvidia-smi. The command I typically use to run the
> simulation is:
>
> *namd2 +p8 .... .conf | tee .... .log *
>
> Do I need to specify anything extra in the command? or in my configuration
> file. Thanks.
>
>
>
> With best regards
>
>
> --
>
> *Roshan Shrestha*
>
> M.Sc (Physics)
>
> Central Department of Physics, Tribhuvan University
> Kathmandu, Nepal
>
>
>
>
>
>
>

-- 
Giacomo Fiorin
Associate Professor of Research, Temple University, Philadelphia, PA
Research collaborator, National Institutes of Health, Bethesda, MD
http://goo.gl/Q3TBQU
https://github.com/giacomofiorin

This archive was generated by hypermail 2.1.6 : Thu Dec 31 2020 - 23:17:13 CST