Folding Forum

I was looking for different views on what the barriers of entry were for introducing FPGA's into folding@home. In the post's I've seen, FPGA's get dismissed too readily because of the GPU's superior floating point capability. However, with the announcements from both Altera and Xilinx moving into 3D transistor technology, I think that their integration needs to be revisited. Some advantages and disadvantages of what I could think of is listed:

Advantages:
more economical to run than GPU
greater ability to accelerate functions
reconfigurable based on working core

Disadvantages:
significantly slower clock rate than GPU (currently)
expensive integration 'cause of FPGA costs and board hardware
not readily applicable

The problem for folding is not GPU's superior FP capability over an FPGA array. The problem is that PG has a very limited programming staff and the ROI of developing a client will likely be far to low to be worth it. There are too few cards/machines with an FPGA in them and it costs a lot to develop a client from scratch so it is unlikely to happen. If you look at history, It took many years of people complaining for them to add a Linux client and many more to add a GPU client to it. I doubt that the numbers of FPGA users compare to Linux users. There have been numerous requests for clients to be made for specialty high-performance FP hardware like Sun or PhysX (Before it was converted to a software standard by Nvidia) and they have always been declined based on that argument.

P.S.

I'm not trying to speak for PG but there is a history here of people requesting new clients for specialty hardware. If PG disagrees then someone will undoubtedly speak up. They do monitor these forums.

We frequently look into FPGA's and they have yet to make sense price/performance-wise. We'll keep an eye on it.