Hardware Question

bmandt

Dear all,

thank you very much for this great piece of software.

To speed up my simulation I would like to know, which of the following cards would be the better choice:
1. ASUS GTX-TITAN-6GD5; 2688 Steam cores; 1000 Euro
2. PNY Quadro 6000; 448 Steam cores, 4000 Euro

Thank you very much for your reply.

Best regards,

Benjamin

SLongshaw

Hi,

The GTX TITAN is a top end Kepler based GPU while the Quadro 6000 is a top end Fermi base GPU.

The Kepler GPU is much newer and significantly faster than the Fermi architecture and the GTX TITAN is nearly as fast a Kepler based GPU as you can buy.

If you compare a Quadro K6000 to the GTX TITAN then the difference between the two will be smaller as the K6000 is Kepler based also, but the GTX TITAN would still the faster GPU for CUDA processing.

In short: unless you are buying a Tesla K40 then the GTX TITAN is about as good as you can get in terms of a GPU for CUDA at this moment.

Best regards.

taffo

I use an eVGA GTX Titan with DualSPHysics and the improvement in terms of computational speed with respect to my previous nVIDIA Quadro 4000 is quite impressive. The Titan is a monster card, I would say go for it, considering also the difference in price.

bmandt

Hi,

thank you very much for your comments.

Makes it sense to purchase 2 cards for parallel useage? Is running dualsphysics on two cards simultaneous (same case) possible?

Best regards,

Benjamin

SLongshaw

Hello Banjamin,

There is a multi-GPU version of DualSPHysics currently in development. At the moment it is being distributed as a Beta compiled only version at request only. In the future it will be released as per the current single-GPU version.

If you have two GPUs in your machine then you will be able to run two cases simultaneously, but not utilise both GPUs to handle one case.

I would suggest however that you have at least two GPUs in a CUDA processing workstation, one slightly less powerful for graphics output and one purely for CUDA as not having a monitor attached to a GPU increases the amount of initial available resources it has, resulting in slightly faster operation.

Best regards