Friday, June 20, 2008

Nvidia Reveals Two Flagship Graphics Processing Units

Nvidia Corp. on Monday formally unveiled two new flagship graphics processing units (GPUs) that bring performance of PC gaming to new heights in terms of performance. But while the new chips offer generally higher performance compared to predecessors and sport unprecedented amount of transistors, they sport no improvements in terms of feature-set.

The new Nvidia GeForce GTX 200-series graphics cards are based on the code-named G200 graphics processor that features 240 unified shader processing processors (SPs, stream processors) , 80 texture units (TUs, 80 texture address, 80 texture filtering units), 32 render back ends (RBEs) as well as 512-bit memory controller. The fresh breed of graphics processors also features double-precision floating point support, re-designed thread dispatch processor, PCI Express 2.0 support and some other tweaks. Like the predecessors, the GTX 200-series is DirectX10 compatible and features hardware acceleration and post-processing of high-definition video playback as well as other capabilities. The new graphics chip consists of 1.4 billion transistors and is manufactured using 65nm process technology at Taiwan Semiconductor Manufacturing Corp. Since the developer could not fit output interface into the G200 chip possibly due to transistor budget, graphics boards powered by the GPU will be equipped with a new version of the NV IO controller that supports D-Sub, dual-link DVI, DisplayPort, TV-Out and other interfaces.

Nvidia said that thanks to software advancements the new GeForce GTX 200-series graphics processors will support H.264 video encoding, Stanford University’s distributed computing computational program Folding@Home as well as will eventually support processing of certain physics effects created using Ageia PhysX tools.

Unfortunately, Nvidia decided not to implement any new hardware features: the new GeForce GTX 200-series graphics processors do not support DirectX 10.1/shader model 4.1, any new methods of full-scene antialiasing or any other features. Still, Nvidia claims, with tangibly improved processing power from the previous-generation flagship chip – from 128 to 240 SPs – new video games will ultimately look better.

Initially Nvidia and its add-in-board partners will sell two versions of the GeForce GTX 200: the premium-class GTX 280 with 240 SPs, 80 TUs, 32 RBEs and 1GB of GDDR3 memory connected using 512-bit interface as well as high-end GTX 260 with 192 SPs, 64 TUs, 28 RBEs and 896MB of GDDR3 memory connected using 448-bit interface. The GeForce GTX 280 will have GPU clocked at 602MHz, SPs operating at 1296MHz and memory working at 2214MHz, whereas the GeForce GTX 260 will be clocked at 576MHz, 1242MHz and 1998MHz, respectively.

Graphics cards featuring GeForce GTX 280 GPUs will be available almost immediately from add-in card manufacturers, such as: Asustek Computer, BFG, Emtek, EVGA, Gainward, Galaxy, Gigabyte, Innovision, Leadtek, MicroStar International, Palit, PC Partner/Zotac, PNY, Point of View, Unika, Unitek, and XFX. Graphics cards featuring GeForce GTX 260 GPUs will be available starting on Thursday, June 26, 2008. Suggested retail pricing for the GeForce GTX 280 and GeForce GTX 260 GPUs are $649 and $399, respectively.

No comments:

Post a Comment