Re: [AMBER] some problems when using pmemd.cuda

From: Jason Swails <>
Date: Tue, 30 Sep 2014 07:11:32 -0400

On Mon, Sep 29, 2014 at 10:33 PM, <>

> Dear Jason M. Swails:
> Thanks very much!
> I use AMBER12 and AMBERTOOLS12. The second GPU wokrs. I submit a 20ns job
> to the second GPU. It can run the job for short time (maybe stopped at 2ns
> or 3ns) without any error message. May it be the problem of the compiler?

​No. It's most likely a problem with the GPU.

> ​​
> I wonder whether there are some methods to submit 2 or more jobs to one
> GPU without evident speed
> ​​
> decrease. One job just cost 500M of the total 5000M GPU memory. This may
> be a little waste of the
> ​​
> computing resources.

​No. You are equating total RAM with "computational power" and that is
just not true (as I said before). If a program makes efficient use of the
clock cycles of ANY processor (be it a GPU or a CPU) then forcing that
program to share clock cycles with another running program will ALWAYS hurt

Like I said before, total RAM has almost no impact on computational
performance -- it's the memory bus speed, processor speed, and available
instruction set that actually matters when it comes to performance.


Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Tue Sep 30 2014 - 04:30:02 PDT
Custom Search