My first thought when reading about this is I know NVIDIA makes rack mounted devices with their "Tesla" GPU for HPC applications like this. I would feel more confident with hardware designed for what the Airforce is trying to do and it seams like the Tesla servers would have a lot less wasted hardware and may be more efficient in a server environment.<p>So I googled it and found out IBM already moved from Cell to Tesla last May.<p><a href="http://arstechnica.com/business/news/2010/05/ibm-cheats-on-cell-with-nvidia-tesla-for-servers.ars" rel="nofollow">http://arstechnica.com/business/news/2010/05/ibm-cheats-on-c...</a><p>I'm also curious how AMD's new APU will affect this market.<p><a href="http://news.ycombinator.com/item?id=2067057" rel="nofollow">http://news.ycombinator.com/item?id=2067057</a>
what a waste. They could have bought IBM BladeCenter solution with the same processors, but with Infiniband and more memory.. instead they have a big toy.