Re: NAMD-CUDA GPU workload

From: David Hardy (dhardy_at_ks.uiuc.edu)
Date: Mon Apr 02 2018 - 11:33:06 CDT

Dear Yu,

Part of the problem is that NAMD 2.12, to balance your fast GPU, requires a lot of CPU support, well beyond an AMD 8-core CPU.

We will soon (within the next week or two) be releasing the first beta of NAMD 2.13. The new release will allow you to offload all of the force calculation to the GPU, which should increase your GPU usage.

Best regards,
Dave

--
David J. Hardy, Ph.D.
Beckman Institute
University of Illinois at Urbana-Champaign
405 N. Mathews Ave., Urbana, IL 61801
dhardy_at_ks.uiuc.edu, http://www.ks.uiuc.edu/~dhardy/
> On Apr 2, 2018, at 10:11 AM, Yu Zhou <zylp2006_at_gmail.com> wrote:
> 
> Dear NAMD users,
> 
>  
> I was running simulation on an 70,000-atom system of an ion channel (KcsA) using NAMD 2.12 for Win64-multicore-CUDA on a desktop with an AMD 8-core CPU and a GTX-1080 graphic card by:
> 
> "NAMD2 +p8 kcsa.conf > kcsa.log". I noticed that the workload of GPU was only ~20% during the whole simulation process. I am wondering if there was any way I could increase the workload of GPU to accelerate simulation.
> 
>  
> Thanks for your help.
> 
>  
> Yu Zhou
> 
> Dept. of Anesthesiology
> 
> WUSM
> 
> 

This archive was generated by hypermail 2.1.6 : Sat Dec 07 2019 - 23:19:42 CST