GPU parallel computing
1 view (last 30 days)
Hi I am looking to buy a new pc to help speed up computations. I read a bit about the gpu support in the Parallel Computing Toolbox. May I ask if anyone has experience of using their GPU for calculations and if it has sped it up?
Additionally could anybody let me know if the GeForce GTX 590 would work?
Malcolm Lidierth on 27 Apr 2011
Edited: Malcolm Lidierth on 20 Aug 2012
The NVidia card needs a "Compute Capability" of 1.3 or higher (supports IEEE double precision) for the Parallel Computing Toolbox. You might also look at Accelereyes' Jacket
The speed up depends on the application - you are adding an overhead of transferring data on the bus to the GPU, which is slow but worth it if you then do a lot of processing with those data.
Jason Ross on 29 Apr 2011
This question has been answered many times on the more traditional "hardware" sites (Tom's Hardware, etc), as well as on the nVidia discussion boards. The people who maintain the pages at Wikipedia also do a fantastic job of summarizing the specs of the numerous nVidia cards, as well.
The best I can come up with based on comparing these cards is "it depends on what you are looking to do". Just like any other design / purchasing decision, there are a lot of factors to consider that only you can determine the answer to. Broadly speaking, the GeForce/GTX line is aimed at the gamer market, the Quadro at professional desktop applications, and the Tesla at compute-intensive work. While they do share common components and architecture, there are differences among the cards, and there is not one "right" answer.