NVIDIA recently unveiled its desktop supercomputer Tesla C1060 whose technology is already being used in bigger systems. NVIDIA has namely contributed no less than 170 Tesla S1070 U1 rack systems to a supercomputer at Tokyo Institute of Technology. The supercomputer is called TSUBAME and its the first using a Tesla-based cluster. The system earlier used a Sun Fire cluster, alternating between Opteron and Xeon processors. After the upgrade to NVIDIA’s Tesla cluster TSUBAME was clocked at a maximal performance of 162 TFLOPS. That makes it the 29th fastest computer on the planet.
Exactly how much that comes from the NVIDIA Tesla cluster isn’t told by the story. Still, this is a great leap forward by NVIDIA and GPGPU-based (General Purpose Graphic Processing Unit) supercomputers as a whole. We can most likely count on more supercomputers sporting GPGPU technology appearing on the Top500 list in the future.
