Dear All,
Santa has arrived early for us and we have a couple of GTX titans in the
lab. So here are the GPU benchmark results:
JAC_PRODUCTION_NVE - 23,558 atoms PME
-------------------------------------
1 x TITAN GTX: ns/day = 107.76 seconds/ns = 801.78
JAC_PRODUCTION_NPT - 23,558 atoms PME
-------------------------------------
1 x TITAN GTX: ns/day = 111.34 seconds/ns = 775.99
FACTOR_IX_PRODUCTION_NVE - 90,906 atoms PME
-------------------------------------------
1 x TITAN GTX: ns/day = 32.54 seconds/ns = 2655.14
FACTOR_IX_PRODUCTION_NPT - 90,906 atoms PME
-------------------------------------------
1 x TITAN GTX: ns/day = 26.20 seconds/ns = 3297.82
CELLULOSE_PRODUCTION_NVE - 408,609 atoms PME
--------------------------------------------
1 x TITAN GTX: ns/day = 7.85 seconds/ns = 11011.91
CELLULOSE_PRODUCTION_NPT - 408,609 atoms PME
--------------------------------------------
1 x TITAN GTX: ns/day = 6.50 seconds/ns = 13290.12
TRPCAGE_PRODUCTION - 304 atoms GB
---------------------------------
1 x TITAN GTX: ns/day = 637.66 seconds/ns = 135.50
MYOGLOBIN_PRODUCTION - 2,492 atoms GB
-------------------------------------
1 x TITAN GTX: ns/day = 211.49 seconds/ns = 408.53
NUCLEOSOME_PRODUCTION - 25,095 atoms GB
---------------------------------------
1 x TITAN GTX: ns/day = 3.90 seconds/ns = 22172.74
The cards we have are Gigabyte GV-NTITAN-6GD's. Hope this will help people
in their decision making processes
Cheers
Ian
Women love us for our defects. If we have enough of them, they will
forgive us everything, even our intellects.
Oscar Wilde,
--
Dr Ian R Gould, FRSC.
Reader in Computational Chemical Biology
Department of Chemistry
Imperial College London
Exhibition Road
London
SW7 2AY
E-mail i.gould.imperial.ac.uk
http://www3.imperial.ac.uk/people/i.gould
Tel +44 (0)207 594 5809
On 28/02/2013 08:41, "Gonzalo Jimenez" <gonzalojimenezoses.gmail.com>
wrote:
>Thanks a lot, Ross, that really helped a lot!
>I cannot wait to see you real-life, more trustable benchmark results with
>the GTX Titan and Amber12. I will make a final decision then, but now is
>99%
>in favor of the GTX 680s.
>Due to technical reasons, we cannot use more than 700W power supplies, so
>we
>usually put 2 GTX n80 in a AMD-based linux box.
>Best regards,
>
>Gonzalo
>
>-----Mensaje original-----
>From: Ross Walker
>Sent: Thursday, February 28, 2013 12:10 AM
>To: AMBER Mailing List
>Subject: Re: [AMBER] GTX 680 vs GTX Titan
>
>>Dear all,
>>
>>We are going to expand our GTX 580-based cluster and we were thinking
>>about buying the Œold¹ GTX 680¹s, but it seems that the Titans are going
>>to be unleashed soon, and you guys are talking really good things about
>>them... We need to spend the money pretty soon, so my questions are:
>>
>>- Is there any chance that we can buy affordable GTX Titans in the next 2
>>weeks or so?
>
>Unfortunately I have to say the answer is no.
>
>>- Which models do you guys recommend?
>
>For GTX680 probably the following EVGA card will give you the best in
>terms of stability and performance. In my experience the eVGA cards tend
>to be pretty solid.
>
>http://tinyurl.com/d82lq8d
>
>
>These are $549 each plus tax. You can get over clocked versions as well
>like the + version of the above or the classified version but they tend to
>cost more money for only a fractional percentage improvement in speed.
>
>>- Which is the expected price for one of these things? Is it worth
>>compared to GTX 680?
>
>For the GTX-Titan - Between $999 and $1199 + tax. So just under double the
>price of a GTX680. Right now I am still waiting on access from NVIDIA. I
>hope to have real benchmarks in a couple of days but we estimate about 115
>to 120ns/day for the JAC NVE benchmark. Compared with around 72.5ns/day
>for the GTX680. You can put 4 of either in a standard box (with a 1400W+
>power supply) so you really just need to do the math on whether it is
>worth it for you. It is hard to estimate the value of being able to run a
>single simulation faster since peoples opinions differ wildly on this.
>Just look at people buying Intel chips. They'll pay 150% more in price for
>a 4.3% boost in clock speed. CRAZY! But yes 120ns/day on 1 card would be
>nice. The other thing of course is whether the GTX680 price will come down
>with wide availability of the GTX-Titan. I don't know the answer to that
>but previous experience with NVIDIA graphics cards suggests that is
>unlikely it will just slowly get phased out. The GTX-Titan price probably
>will come with time though but it will quite a while.
>
>So I guess if you have to spend the money in 2 weeks then GTX680 is really
>your only option (690 is the same price as the Titan). If you can wait 6
>weeks or so then it is a coin toss between the 680 and the titan depending
>on how many nodes you have to put these cards in. Also note that right now
>the Titan is an unknown quantity since we haven't actually had a chance to
>run on a real one yet.
>
>Hope that helps
>All the best
>Ross
>
>/\
>\/
>|\oss Walker
>
>---------------------------------------------------------
>| Assistant Research Professor |
>| San Diego Supercomputer Center |
>| Adjunct Assistant 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 Thu Feb 28 2013 - 03:00:03 PST
