Which cuda solution to buy in the next 2 months?

My group is making the plunge into GPU computing and have decided to use Cuda. The newer features provided by Fermi, including double point support make it particularly attractive. The particular problem we have is that the grant money we have earmarked for this is of the “spend-it-or-lose-it” variety. It expires in the next couple months (don’t have the exact date yet). We have been waiting for the Tesla Fermi options to become available for sale, but that may not be in our window now. While there has been talk on these boards covering the cost benefit debate for the Tesla/Graphics Card pre-Fermi boards, I was wondering if we had enough cost/performance information to predict how things will shake out for Fermi Teslas/Quadros? If I have to spend the money now should I go with a Fermi based graphics card if I can find one? Should I just buy the non-Fermi equipment? Those in the know, please help!

Thanks!
-David

This is a producer(GNT)-consumer (NV) problem.

This can be solved in 2 ways. One is to negotiate with the consumer and the other is with the producer.

  1. See if NVIDIA come up with TESLA-FERMI exchange offer. So, you can buy some TESLA C1060s today, and re-convert it to FERMI at a later date. Try to forge a deal with NV on that (depending on the number of TESLAs u buy - u might be in a position to seal a deal) (I vageuly remember such a deal being floated by NV. Not too sure if still valid… Check out)

  2. Deposit the money in fixed deposit and use that interest to buy a FERMI at a later date and give back the principal amount to the GNT provider (OR) Try negotiating an exotic financial option with your grant provider to make this happen…