Re: [AMBER] pmemd.cuda on an Amber Certified High-End workstation

From: Li, Wen <wl2290.cumc.columbia.edu>
Date: Thu, 20 Apr 2017 19:52:43 +0000

Dear Dr. Case and Dr. Walker,

Thank you very much for your kind help! I did all you advised, and pmemd.cuda run well for the second heating with reduced constraints after the first step.

Best wishes,
Wen

________________________________________
From: David Case [david.case.rutgers.edu]
Sent: Tuesday, April 18, 2017 1:29 PM
To: AMBER Mailing List
Subject: Re: [AMBER] pmemd.cuda on an Amber Certified High-End workstation

On Tue, Apr 18, 2017, Li, Wen wrote:
>
> I am running pmemd.cuda on an Amber Certified High-End workstation with
> a 24GB P6000 Quadro card. The solvated system includes about 800,000
> atoms. Running minimization worked, but running heating got the error
> message as the following. I am wondering if the molecular system is too
> large to use this workstation.
>
> Error: an illegal memory access was encountered launching kernel kC
> learForces cudaFree GpuBuffer::Deallocate failed an illegal memory
> access was encountered

Did you get any output? Try a short run with ntpr=1 to try to get more
information: look for high temperatures or rapidly changing energies. This
would be indicative of problems with your system, not with the GPU.

You could also try a short equilibration on a CPU, then try to continue on
the GPU.

....dac


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Apr 20 2017 - 13:00:02 PDT
Custom Search