Asus has unveiled a mini-supercomputer based on Nvidia graphics chips and a 3.33GHz Xeon.
The ESC 1000 (pictured) can use a combination of traditional microprocessors and graphics chips to attain speeds of over 1.1 teraflops, according to Asus.
The mini-supercomputer, which was developed in conjunction with graphics tech company Nvidia, can use an Nvidia Quadro FX5800 graphics card, combined with three Nvidia 960 graphics processing cores in three Tesla c1060 computing processors. In the ESC 1000 Ultimate model, these cards are combined with a more traditional microprocessor, a 3.33GHz Intel LGA1366 Xeon W3580.
Using this combination of cards, the ESC 1000 can reach up to 1.104 teraflops. Asus said it had used single precision benchmarking to determine the processor speed.
'Flops', or floating point operations per second, is a measurement used for processor speed. One teraflop is equivalent to 1012 flops. The world's most powerful supercomputers, including IBM's Roadrunner, run at much higher speeds, are measured in petaflops, or 1015 flops.
Talkback
so by the looks of things the cost of creating a local super computer have fallen then, mind you where this stands up against renting a slot on virtual cloud processing unit I'm not sure if it's intended to or not.
Came across this article the other day that might be of some interest to people, by the looks of things both Intel & AMD are going to have some competition soon, which not a bad thing.
http://www.eweek.com/c/a/IT-Infrastructure/Tilera-Talks-100Core-Processor-376613/