Re: [AMBER] Does anyone know a screening program of hardware validity for GeForce GTX board?

From: Hiromichi Tsurui <tsurui.juntendo.ac.jp>
Date: Thu, 9 Jun 2011 10:11:44 +0900

Thank you Scott,

The run was performed without constraints.
This is the prod.in file I used for the run.
http://enzyme.fbb.msu.ru/Tutorials/Tutorial_1/Sec_5.htm
I changed cut=10.0 to cut=8.0. The sample is a crystal structure of p-MHC
from PDB.

Hiro



From: "Scott Le Grand" <varelse2005.gmail.com>
To: "AMBER Mailing List" <amber.ambermd.org>
Sent: Thursday, June 09, 2011 2:26 AM
Subject: Re: [AMBER] Does anyone know a screening program of hardware
validity for GeForce GTX board?


This is an NTP simulation or an NTP simulation with constraints?

If so, it's likely to be a bug fixed in the soon-to-be-released patch. As
for the definition of soon, that's Ross's department...

2011/6/8 Hiromichi Tsurui <tsurui.juntendo.ac.jp>

> Hi all,I ran MDs of about 60000 atoms using pmemd.cuda_SPDP with MSIís
> GTX580 1.5GB and other manufacturer's 3GB board. MDs with MSIís board
> always ran through 10 ns, 5000000 steps completely. However, otherís board
> often crashes presenting error message saying ďmax pairlist cutoff must be
> less than unit cell max sphere radiusĒ. Of course I patched bugfix all. I
> think the difference of memory 1.5GB and 3.0GB does not affect on this
> case.
> I am afraid a slight hardware flaw in phantom caused this error. Does
> anyone
> know a program that detects detailed failure in GTX board? Or give any
> suggestion?Hiro
> _______________________________________________
> 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


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jun 08 2011 - 18:30:02 PDT
Custom Search