Re: [AMBER] Performance of RuBisCo (74, 916) atoms on NVIDIA Tesla K40 GPU

From: Gerald Monard <Gerald.Monard.univ-lorraine.fr>
Date: Wed, 13 Aug 2014 21:36:08 +0200

Hello,

Please have a look at section 18.6.7 "Considerations for Maximizing GPU
Performance" of the Amber14 manual. You should set ntpr and ntwx to
values much greater than the one you have (at least 1000); you should
also prefer ntt=1 or 2 over ntt=3 for better performance.

Gerald.


On 08/12/2014 09:03 AM, Amit Harode wrote:
> Dear Dr. Ross,
>
> Thank you for your response!
>
> I was using DPFP precision model because of its higher accuracy. I am using the following "mdin" file:
>
> &cntrl
> irest=0,
> ntx=1,
> igb=2,
> rgbmax=14.0,
> cut=9999.0,
> ntpr=10,
> ntwx=100,
> nstlim = 500,
> dt=0.001,
> ntt=3,
> tempi=300.0,
> temp0=300.0,
> ntb=0,
> nscm = 0,
> &end
>
> So one option to improve performance is to use SPFP precision. Apart from the precision model are there any other knobs that can improve performance? I am using the latest release of Amber (AMBER14) and Tesla K40 GPU. It is a Generalized-Born (GB) implicit solvent model calculation for RuBisCO (PDB ID: 1IR2, is a 74,916 atom protein ) (Ref : David E.Tanner, James C. Phillips, and Klaus Schulten, J Chem Theory Comput. Jul 10, 2012; 8(7): 2521&#8211;2530 ).
>
> Warm Regards
> Amit N. Harode
> TCS INDIA
>
> -----Ross Walker <ross.rosswalker.co.uk> wrote: -----
> To: AMBER Mailing List <amber.ambermd.org>
> From: Ross Walker <ross.rosswalker.co.uk>
> Date: 08/11/2014 10:40PM
> Subject: Re: [AMBER] Performance of RuBisCo (74, 916) atoms on NVIDIA Tesla K40 GPU
>
>
> Hi Amit,
>
> Firstly the DPFP precision model is meant for debugging only, hence
> performance is very bad, especially on GeForce cards. Use the default SPFP
> precision model and you should get much better performance.
>
> Beyond that without knowing the details of your simulation, your mdin
> file, total number of atoms (including solvent), AMBER version and the GPU
> model it is hard to make any additional suggestions.
>
> All the best
> Ross
>
>
> On 8/11/14, 5:34 AM, "Amit Harode" <amit.harode.tcs.com> wrote:
>
>> Hello all!
>>
>> I am using AMBER-14 on NVIDIA Tesla K40 GPU. I got performance of 0.04
>> ns/day with RuBisCO (PDB ID 1IR2) using DPFP precision (Double Precision
>> Fixed Precision).
>> Has anyone got better performance? Are there any settings that will help
>> improve performance?
>>
>> RuBisCO (PDB ID 1IR2) is a 74,916 atom protein system (Ref : David E.
>> Tanner, James C. Phillips, and Klaus Schulten, J Chem Theory Comput. Jul
>> 10, 2012; 8(7): 2521&#8211;2530 )
>>
>>
>> Thanks & Regards
>> Amit N Harode
>> TCS INDIA
>>
>>
>> =====-----=====-----=====
>> Notice: The information contained in this e-mail
>> message and/or attachments to it may contain
>> confidential or privileged information. If you are
>> not the intended recipient, any dissemination, use,
>> review, distribution, printing or copying of the
>> information contained in this e-mail message
>> and/or attachments to it are strictly prohibited. If
>> you have received this communication in error,
>> please notify us by reply e-mail or telephone and
>> immediately and permanently delete the message
>> and any attachments. Thank you
>>
>>
>> _______________________________________________
>> 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
>

-- 
____________________________________________________________________________
  Prof. Gerald MONARD
  SRSMC, Université de Lorraine, CNRS
  Boulevard des Aiguillettes B.P. 70239
  F-54506 Vandoeuvre-les-Nancy, FRANCE
  e-mail : Gerald.Monard.univ-lorraine.fr
  tel.   : +33 (0)383.684.381
  fax    : +33 (0)383.684.371
  web    : http://www.monard.info
____________________________________________________________________________
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 13 2014 - 13:00:03 PDT
Custom Search