Re: [AMBER] Amber CUDA calcualtion on GeForce GTX 590 ?

From: Scott Le Grand <>
Date: Wed, 30 Mar 2011 11:37:21 -0700

GTX 590s are clocked lower than GTX 580s approaching 1.6 580s worth of FLOPS or so...

-----Original Message-----
From: Ross Walker []
Sent: Wednesday, March 30, 2011 11:30
To: 'AMBER Mailing List'
Subject: Re: [AMBER] Amber CUDA calcualtion on GeForce GTX 590 ?

Hi Marek,

Yes, these cards should work in exactly the same way as the GTX295's did.
That is they are effectively two completely independ GPUs on the same PCI-E
bus. Consider them to just be essentially 2 GX580s in one of those break out
boxes and note that the effective PCI bandwidth between the two GPUs is
effectively 8x. Hence you can run 2 single GPU jobs at one, one on each GPU
ID and they should run more or less at full single GPU speed. In parallel
the scaling will be slightly worse than if you bougth 2 GTX580 cards and put
them in side by side x16 slots.

Hope that helps.

All the best

> -----Original Message-----
> From: Marek Maly []
> Sent: Wednesday, March 30, 2011 2:40 AM
> To:
> Subject: [AMBER] Amber CUDA calcualtion on GeForce GTX 590 ?
> Hello all,
> I am just wondering if someone already tried to run Amber cuda
> calculations
> on GeForce GTX 590.
> The advantage of this card is double memory and double number of cores
> comparing to GTX 480.
> It seems to me that they just put two chips from GTX 480 in one
> single GPU so I assume compatibility with amber as it is compatible
> with
> 480 version. What I am just a little afraid are cooling issues of such
> card
> especially for long runs ...
> Any experiences, comments appreciated.
> Best wishes,
> Marek
> _______________________________________________
> AMBER mailing list

AMBER mailing list
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.

AMBER mailing list
Received on Wed Mar 30 2011 - 12:00:03 PDT
Custom Search