Geforce 8900GTX and 8950GX2 details listed

With 8900 GTS and 8900 GS at 80 nanometres

By Fuad Abazovic: Thursday 15 February 2007, 13:10
JUST A DAY after we uncovered the existence of Nvidia's Geforce 8950 GTX and Geforce 8900 GTX, our friends in Taiwan confirmed the news. You can read our original stories here and here.

The Geforce 8950 GX2 is a dual-chip card based on a new 80 nanometre G80 chip, probably codenamed something else. Both GX2 GPUs are clocked at 550MHz and the difference is GDDR4. The card comes with 2x512MB of 256-bit GDDR4 memory clocked at 2000MHz. The card has 96 Shaders, per chip. It will be priced at $600.

The second in Nvidia's spring line-up is the Geforce 8900 GTX clocked at 700MHz GPU and 2200MHz memory. The chip still has a 384-bit memory interface and comes with 768MB of memory. The card uses a new 80 nanometre chip and has 128 Shaders. Compared to the Radeon X2800XTX it will end up shorter on clock and memory interface. You can compare them here.

The Geforce 8900 GTX is priced at $550.

Meanwhile, the Geforce 8900 GTS is a new card clocked at 600MHz GPU with 2000MHz GDDR4 memory. It supports the 320-bit memory controller and comes with 640Mb of memory. This card should cost $500 and it is using the 80 nanometre chip. This card will still have 128 Shader units.

The current king of the crop, the 8800 GTX will drop in price to $450, while the Geforce 8800 GTS loaded with 640MB of memory stays up in the $400 price range.

Nvidia has two more 80 nanometre cards. The Geforce 8900GT with 600MHz core and with a 256-bit memory interface comes with 512MB of 1800MHz GDDR3 memory. It has 96 Shaders and is built on a 80 nanometre process and will cost $400.

The 8900 GS will be the cheapest G80-based card. The 80 nanometre based beast is clocked at 550MHz core and 1600MHZ memory. The card has the 256-bit memory controller and comes with 256 or 512MB of GDDR3 memory. It also has 96 Shaders. The 256MB version will cost $200 while 512MB incarnation will end up at around the $250 price mark.

The apparently-leaked document mentions G84 and G86 chips but we have some details to confirm. The original story with document is here. µ

The technology is compatible with 8800-series graphics cards, which are estimated to achieve a peak performance of about 520 GFlops, allowing consumers to build a teraflop system using Nvidia's SLI bridge. Andy Keane, general manager of professional products at Nvidia, told TG Daily that the CUDA kit is available to anyone interested, but it will take the knowledge of an "educated programmer" to tap the potential of the card. "Casual programming know-how won't be enough," Keane said. The technology is based on C-language and comes with what Nvidia claims is the first C compiler for the GPU.