Dear Dac,
thanks for the tip. It was infact not minimized adequately. I took some time to minimize it carefully and then heat it up carefully ( NVT with dtt = 0.1 , gamma_linc = 15.0, nstlim = 50000) then in subsequent steps reduced the gamma_linc and increased is the dtt as i proceeeded to NPT .
I ended up doing 6 short equilibrations steps ( NVT --> NPT with decreasing restraints --> NPT with not restraints) . not sure if it was a overkill.
These measures fixed my issues , thanks a lot for pointing me in the right direction.
thanks
amith
________________________________
From: David A Case <david.case.rutgers.edu>
Sent: Wednesday, October 20, 2021 9:17 AM
To: AMBER Mailing List <amber.ambermd.org>
Subject: Re: [AMBER] cudaMemcpy GpuBuffer::Download failed an illegal memory access was encountered
On Tue, Oct 19, 2021, Rangarajan, Amith wrote:
>
>I am encountering this error frequently in a non-specific way ( same system
>-error on only some replicas, corrects upon retry) . I have checked the
>systems in VMD and there seems to be nothing wrong with them. It happens
>mostly during the beginning of the MD run and i dont find any other error
>messages.
The actual error message says nothing useful, other than that an error
occurred. If it happens at the beginning of MD, you probably have some bad
contacts or other problems, even with a partially minimized structure.
Do a short run (just a 100 MD steps) with ntpr=1, and see if you can learn
anything.
You may need to do more minimization (with the CPU), or run the short MD
described above on a CPU.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
https://urldefense.com/v3/__http://lists.ambermd.org/mailman/listinfo/amber__;!!LQC6Cpwp!-DBhlzdE8PukoO0otXycBL75re7eI7VzQCqGFqo4EgYxpHTr40L4OecVTdYMsJn59JRATQ$
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Oct 20 2021 - 10:00:03 PDT
