GPU card accelerates signal/image processing
Based on the new Kepler GPU architecture with 1,536 cores/GPU, the Tesla K10 accelerator board provides 4.58 TFLOPS of single-precision floating-point processing and has 320 Gbytes/s of memory bandwidth. The card uses two Kepler GPU processors and 8 Gbytes of memory.
The K10 is said to enable 1 Petaflop performance server systems in just 10 racks and taking only 400 kW. It provides three times the performance/watt compared to the previous card. The Kepler GPU allows 32 simultaneous MPI tasks, while the previous Fermi architecture ran just one. (Price not available June delivery.)
By Jim Harrison
NVIDIA , Santa Clara , CA
Sales 408-486-2000
www.nvidia.com
Advertisement
Learn more about NVIDIA
