4

I am looking into GPU computing and I can't figure out what the technical / performance differences are between a NVIDIA Quadro 6000 and a NVIDIA Tesla C2075 graphics card. They both have 6GB of RAM and the same number of computing cores. So what's the difference? I want to do CUDA computations with the card.

EDIT: Please, if Nvidia says that card X is good for climate calculations, card y is great for seismic processing, than this is nothing but PR. There is no graphic card made for climate calculations. A card is either good for single or double precision computing, or for FFTs etc. And that's exactly my questions here: what are the technical differences and for what kind of computations should I expect to get faster results on one card vs the other.

memyself
  • 11,907
  • 14
  • 61
  • 102

3 Answers3

5

The biggest hardware difference is that Tesla cards have ECC memory, which is important if you're doing long computations and you want to be able to believe the results.

Jonathan Dursi
  • 50,107
  • 9
  • 127
  • 158
0

Tesla C2075 has a Dual-Link DVI-I and a Maximum Display Resolution of 1600x1200

TheGreatPower
  • 195
  • 1
  • 2
0

"NVIDIA® CUDA™ parallel computing architecture is enabled on GeForce®, Quadro®, and Tesla™ products. Whereas GeForce and Quadro are designed for consumer graphics and professional visualization respectively, the Tesla product family is designed ground-up for parallel computing and offers exclusive computing features."

http://www.nvidia.co.uk/page/why-choose-tesla.html

Udrian
  • 76
  • 6
  • I would really like to see some specification differences / performance difference. The text you quote doesn't say anything about the WHY or WHAT of the difference. It's just marketing. – memyself Oct 12 '11 at 08:48
  • Well it depends on what you want to do, the Quadro is specialized in visual computing like ray tracing and other 3D graphics problem. The tesla is ideal for other computation like seismic processing, biochemistry simulations, weather and climate modeling, signal processing, computational finance. So it all depends on what type of CUDA computations you want to do – Udrian Oct 12 '11 at 09:02