From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Fri Jun 28 2019 - 02:22:53 CDT
Interpreting pure GPU utilization isn’t really a good option.
What is the timing difference (ns/day) with and without GPU?
Are you using as much CPU cores as under Linux?
Regards
Norman Geist
Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im Auftrag von Pratik Narain Srivastava
Gesendet: Freitag, 28. Juni 2019 09:18
An: namd-l_at_ks.uiuc.edu; Miro Astore <miro.astore_at_gmail.com>
Betreff: Re: namd-l: does any body know why GPU Acceleration not work on Windows?
The counter question is that why on earth are you using something as bad as windows for research work. Just switch to linux and it works like a charm.
On Fri, 28 Jun, 2019, 10:53 AM Miro Astore, <miro.astore_at_gmail.com <mailto:miro.astore_at_gmail.com> > wrote:
I have had this issue as well, I ran into so many issues with vmd and namd on windows that I eventually switched to windows. I am running namd on linux and get about 70% utilisation.
Sorry I couldn't be more helpful.
On Fri, 28 Jun 2019 at 15:15, Milad Lagzian <milad.lagzian_at_gmail.com <mailto:milad.lagzian_at_gmail.com> > wrote:
Dear all,
I ask this question before, but unfortunately, no one answered. I have seen that some other guys have the same problem.
Why GPU acceleration of NAMD_CUDA does not work on windows OS? No matter what system we use, only around 10% of GPU is engaged in a calculation.
There is no problem with the CPU version of NAMD.
All drivers have installed and tested including the latest CUDA toolkit and runtime, GPU and ...
It seems it is the program issue.
Any suggestion?
This archive was generated by hypermail 2.1.6 : Tue Dec 31 2019 - 23:20:50 CST