Tag Archives: supercomputer

ETech 2008

Two talks stuck out today. One was a brief talk from Saul Griffith of Makani Power. What I took away from the talk was the impracticality of many renewable energy sources such as tidal power and wind power. A combination of all renewable options, with an emphasis on solar and a strong conservation element will be required to simply maintain our current consumption, let alone address growing desires from China and other developing countries.

Another standout talk that actually shared a bent on energy consumption, was by Stan Williams of HP Labs. His lab is researching several pieces of the exabyte and zetabyte computing puzzle. To paraphrase Stan, “We can’t use spinning disks in zetabyte computing because we’ll start torquing the earth” and that “using a zetabyte computer to model the earth’s climate will have to include the computer itself as it will be the single largest actor on the earth’s climate.” The tie-in with Saul’s talk is that a zetabyte computer made of modern components would require 3 terawatts of electricity, which I immediately remembered as the total energy that could be harnessed from tidal power before bringing the oceans to a standstill. Of course, we would also have to include that in the climate model.

Does Processor Speed Still Matter?

I would have said no until the release of so many multi-core processors. Now they have swarmed the Top 500 supercomputer list. But not every supercomputer has a cluster design because not every challenge is solved faster on a cluster. Will there continue to be custom-designed high-performance systems or will clusters take over completely? At some point, cluster hardware is so cheap that it might still perform the calculations faster than special-design hardware.

Barcelona Supercomputer