updated 09:55 am EDT, Thu June 21, 2007
NVIDIA today added a third desktop graphics series to its lineup. The Tesla is the first graphics chipset explicitly built for both graphics and high intensity, general-purpose computing. Programs written for the chip maker's CUDA software platform can not only use its 3D performance (drawn from the GeForce 8800 and Quadro FX 5600) but also repurpose the shader processors for advanced math. The massively parallel nature leads to tremendous gains in performance compared to regular CPUs, NVIDIA claims: in science, calculations have seen speed boosts from a relatively small 45X to as much as 415X in processing MRI scans for hospitals. Increases such as this can mean the difference between using a single system and a whole computer cluster to do the same work, the company adds.
Tesla separates itself further from the Quadro through scaling. While usable by itself, the 518-gigaflop Tesla C870 graphics card ($1,500) can be paired with one or more extra cards inside a Linux or Windows XP system to divide the workload; the D870 ($7,500) resembles the Quadro Plex and offers two cards in an external case for over a teraflop of performance with smaller workstations or those those graphics slots are already full. Topping the range is the S870 server ($12,000); fitting into a single 1-unit rackmount body, the system includes four Tesla chips for two-teraflop speed and can be stacked with other servers to provide even more performance in particularly demanding circumstances.
Shipping directly from NVIDIA itself, the C870 and D870 should be available from the company in August. The server's technology is due considerably later and should arrive in late fall.