I agree with Scott here there are lots of variation in the "hand tweaked" clock speeds on various 580's
I have a selection of cards including POV TGT's which go upto 810MhZ, it's only really sane to put one of them per box as the heat they generate is quite spectacular and I would suspect that three in a box would be "unwise". They scale exactly with clock speed as do all the other variants that I have.
Have just taken delivery of a "water cooled" EVGA GTX 580 at 850MHz but as it's exam time here and I have not had a chance to practice my car mechanics and build the machine it's going into :-) I'll let you know my experiences when I've built it. Being in a chemistry department I am wondering if I can siphon off some of the liquid Nitrogen that's pumped around the place and keep things cool that way! I am so ancient I have run code on the nitrogen cooled Cray's of the '80's and 90's so it really is "Back to the Future Folks" thermal issues will dominate the future.
In answer to the NODPTEXTURE question Scott yes I have done that, Ta.
Also people, I have Tesla 2050's which have always been rock stable no failures on them. It's the Aston Martin Vantage vs Suberau Imprezza argument, the latter costs less to buy goes faster if you are prepared to try, but it will blow up faster as its engineering tolerances will not be as good as the Astons. Tesla's come with 3 year warranties here in the UK, some GTX's with as little as 12 months, so it'll be interesting to see what happens over a protracted period of use. For the curious I drive a Honda Insight as this is so crap I cann't get pulled over for speeding and so still retain my licence, the few grey hairs Ross has are from the time he was a grad student with me and sat in the VW at high speeds! I'm sure he'll tell you his tales of fright over a beer or two if you ask him.
On 01/06/2011 15:40, "Scott Le Grand" <varelse2005.gmail.com<mailto:varelse2005.gmail.com>> wrote:
I have the same 580 as Ian. Overall, I'd expect 580s to equivalent to
each other but some have tweaked their clocks and the higher the clock
the faster the AMBER.
That said, a 2070 will crush a 580 in full double-precision mode so
make sure you're happy with mixed-precision before pulling the trigger
on this.
You'll also probably have to uncomment NODPTEXTURE in gputypes.h to
avoid a hardware bug mysteriously not present in 20xx. Did you end up
doing that Ian?
Scott
On Wed, Jun 1, 2011 at 1:54 AM, Gould, Ian R <i.gould.imperial.ac.uk<mailto:i.gould.imperial.ac.uk>> wrote:
What I meant to say in very confusing language is that pmemd CUDA cannot
take advantage of the SLI mode.
Cheers
Ian
On 01/06/2011 09:53, "Gould, Ian R" <i.gould.imperial.ac.uk<mailto:i.gould.imperial.ac.uk>> wrote:
Hi German
I don't have the Zotac, but have 3GB GTX 580 from Palit which work very
well. You cannot use them in SLI-mode or any other GPU card for AMBER.
Cheers
Ian
On 01/06/2011 09:24, "german.erlenkamp.pharmazie.uni-halle.de<mailto:german.erlenkamp.pharmazie.uni-halle.de>"
<german.erlenkamp.pharmazie.uni-halle.de<mailto:german.erlenkamp.pharmazie.uni-halle.de>> wrote:
Hi Amber-Users,
we're planning to buy some Cuda cards for our group.
Do you have experience with a
Zotac GTX 580 3GB
in SLI-mode?
Thanks for your help.
Best regards,
German
--
German Erlenkamp
Institut fuer Pharmazeutische Chemie
Martin-Luther-Universitaet Halle
Wolfgang-Langenbeck-Str.4
06120 Halle/ Saale
E-Mail: german.erlenkamp.pharmazie.uni-halle.de<mailto:german.erlenkamp.pharmazie.uni-halle.de>
Phone: (49)345 - 55 25 194
_______________________________________________
AMBER mailing list
AMBER.ambermd.org<mailto:AMBER.ambermd.org>
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org<mailto:AMBER.ambermd.org>
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org<mailto: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 01 2011 - 08:30:02 PDT
