Thanks for the feedback Divi. This is very useful and I am glad things are
working for you now. So far I've yet to see any problems with the
benchmark suites so it is good to get some variation in the testing -
especially for larger simulations.
One datapoint that is missing right now is parallel. So if you have time
it would be nice if you could rerun your simulation across both GPUs with
mpirun -np 2 - essentially run the 100,000 steps twice one after the other
and see if they come out identical.
Thanks.
All the best
Ross
On 8/7/13 5:53 PM, "Divi/GMAIL" <dvenkatlu.gmail.com> wrote:
> Ross and others:
>
> I tested the new update on my systems: My machine has 2 Titans. Ran
>the
>MD/PME under NVT and NPT conditions on both GPUs simultaneously using the
>same input four times. Results are perfectly reproducible over 100,000
>steps. The system size: 335968 atoms (A protein-protein complex of three
>large multi-domain protein trimer in water)
>
>GPU-1 (NVT)
>test1.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>test2.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>test3.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>test4.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>
>GPU-2 (NVT)
>test1.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>test2.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>test3.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>test4.out: Etot = -802324.4629 EKtot = 212768.3281 EPtot
> = -1015092.7910
>
>
>The results are perfectly reproducible under NPT conditions also (not
>shown). Also, I do not see any performance degradation on latest driver.
>I
>get 7.4 ns/day under NVT on this system which is similar on previous
>drivers.
>
>I have not tested any benchmark data as this might have been done
>extensively by AMBER/NVIDIA team.
>
>My TITAN card: ASUS GTXTITAN-6GD5
>
> Finally, I SEE a return on my investment on TITAN cards.
>
>Sincere thanks to the hardwork by AMBER/NVIDIA team.
>
>Divi
>
>
>
>-----Original Message-----
>From: Ross Walker
>Sent: Tuesday, August 06, 2013 1:41 PM
>To: AMBER Mailing List
>Subject: [AMBER] GTX-Titan / GTX780 Fix
>
>Dear All,
>
>NVIDIA has just released an updated driver that is designed to address the
>issues seen with GTX-Titan and GTX-780 cards. This has been tested
>extensively in house but has yet to be fully tested publicly. I would
>encourage those of you with GTX-Titan and GTX-780 cards to try installing
>this driver and see if it addresses the issues you have been seeing with
>those cards. I'd be interested in your feedback and also reports of
>performance.
>
>The driver is here:
>
>http://www.nvidia.com/object/linux-display-amd64-325.15-driver.html
>
>I recommend using the CUDA 5.0 toolkit. A patch for CUDA 5.5 with AMBER 12
>will be coming shortly.
>
>For those who want a test suite to use the following is useful:
>
>https://dl.dropboxusercontent.com/u/708185/AMBER_780_Validation.tar.gz
>
>All the best
>Ross
>
>/\
>\/
>|\oss Walker
>
>---------------------------------------------------------
>| Associate Research Professor |
>| San Diego Supercomputer Center |
>| Adjunct Associate Professor |
>| Dept. of Chemistry and Biochemistry |
>| University of California San Diego |
>| NVIDIA Fellow |
>| http://www.rosswalker.co.uk | http://www.wmd-lab.org |
>| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
>---------------------------------------------------------
>
>Note: Electronic Mail is not secure, has no guarantee of delivery, may not
>be read every day, and should not be used for urgent or sensitive issues.
>
>
>
>
>
>
>
>_______________________________________________
>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 Aug 07 2013 - 18:30:02 PDT
