2 replies to this topic

[cosmicstring]

I am planning to buy a new graphics card to use the CUDA support and GPU programming (CUDA C++, CUDA Fortran and CUDA support in Maple).

I have two choices:

1. A professional graphics card (PNY Quadro 600 1GB GDDR3)
2. A "normal" (or gamers') card (ASUS ENGTX480/2DI/1536MD5 1536 MB DDR5 384 Bit or MSI N570GTX-M2D12D5 GTX 570 1280 MB GDDR5 320 Bit).

Which one would be more efficient for GPU programming?

Please note that I will not use any programs related to CAD-CAM or 3D design.

[Alexander]

If I really try to remember, I recall the Quadro containing the same Gxx chips as most of the gaming cards available. You will want to buy a Quadro or Tesla if..

• You require <4-6GBs of fast graphics RAM
  
• You require more rugged testing in stability and performance, in regards to double precision computing
  
• More floating point operations per second

There are likely some hard specifications you can look at on their website, for Maple or casual number crunching it would seem a bit silly to me.

Alexander.

[bbqroast]

Mmmm...
A Nividia GeForce would be good as they are very widely used... So the second option of option 2 (I think). I am no GPU programmer but I imagine programming for different GPUs under the same product 'tree' (GeForce for example) would be the same.

I do recommend you use a rendering engine such as OpenGL as it will be readily prepared for hundreds of cards.