updated 10:00 am EDT, Mon June 16, 2008
NVIDIA GeForce GTX 200
NVIDIA today took the wraps from its anticipated GTX 200 series video chipset, introducing its first new architecture since the 8-series appeared in late 2006. All new boards have an architecture which is as much as 50 percent faster than even the top-end 8800 Ultra; the fastest part in the launch, the GTX 280, nearly doubles the amount of effects processors to 240.
All of these are also the first to support NVIDIA's CUDA general-purpose technology that allows supporting apps to use the video chipset for PhysX-based physics in games as well as calculations in other tasks; a GTX 280 is 45 times faster in Folding@Home genetic calculations than a 3GHz Core 2 Quad, NVIDIA claims.
The GTX 280 supplies the full feature set with a 602MHz core clock speed, and 1GB of memory on a 512-bit bus running at 1.1GHz actual speed; the GTX 260 scales the number of processors back to 192 and also dials back the core and memory speeds to 576MHz and 999MHz respectively; it carries an unusual 896MB of video RAM on a 448-bit bus.
All the cards support up to three-way SLI graphics and have full hardware acceleration for most HD video formats. The cards will be available both in pre-assembled systems, starting with a new version of the HP Blackbird 002, as well as from third-party card manufacturers such as ASUS, BFG, eVGA, and XFX. NVIDIA's official recommended prices sit at $399 for the GTX 260 and scale up to $649 for the range-topping GTX 280. Both boards will be ready by June 26th.