Here’s a reconfigurable computing blog of note I just came across. Check out this posting by Amir Hirsh about server power management: Datacenter Power Management: Subverting the Dominant Paradigm.
Last month I attended the Washington Technology Summit where power efficiency and alternative energy were big topics (along with life sciences and advanced materials). During a panel discussion on Green IT, Christian Belady of Microsoft described methods that Microsoft and its competitors use to power and cool the hundreds of thousands of CPUs that now make up a large scale data center.
During the Q&A I couldn’t help raising my hand and asking the obvious question: aren’t we attacking the wrong problem? If most of the general-purpose processors in the data centers are being applied to highly specialized problems (web-search, for example, or streaming video transcoding), they why aren’t we making more use of lower-power reconfigurable hardware?
Christian had sort of a blank look (after all, his job is watching the power meter on the building, not deciding what actually goes in all those server boxes) but another panelist, Don Tilton of Spraycool knew exactly what I was getting at. He took the microphone and opined that “GPU and FPGA accelerators are a potential solution” to the problem.