If I remember correctly, the Cell design most closely resembles that of supercomputers, basically you take a lot of processors, and then join them together to work in parallel, usually using specialized hardware to run these chips in a method differently then designed (i.e. using desktop computer chips and running them not in dual or quad setups but 1,000 chip setups).
This concept is also similar to distributed computing projects (folding@home, seti@home, etc) where you take a lot of "individual" computers and use them all at once to process work faster.
With this said, I think that the cost of the chip will be very expensive because the amount of work that each "cell" can do is not known at this time (until there is a sample that can be benchmarked, all claims are just estimates/marketing). I am very interested in how this chip may change the future of computing but it's hard to say until they have put out working samples that can show off the potential of the design/chip.
|