Nvidia (opens in new tab) 's next flagship graphics chip will be twice as powerful as its existing GeForce 8800 GTX GPU. And it could appear before the end of the year. That's the implication following a recent announcement in which the Californian graphics outfit revealed the first details about the upcoming GPU.
Thought to be codenamed G92, Nvidia says the chip will crank out nearly one trillion floating point operations per second - otherwise known as 1Tflop. That's around twice as many as the GeForce 8800 GTX. Speaking to investors during a recent conference call, Nvidia representative Michael Hara did not reveal any further architectural details.
New card for Christmas?
However, he did confirm the company plans to stick to the product launch cycle introduced with the GeForce 8800 GTX. In practice, that means the new chip is due out before the end of the year, with mid-range parts to follow early in the new year.
Various rumours are currently circulating regarding the finer details of the G92 architecture. The smart money indicates the chip will be closely related to the GeForce 8800 but will offer higher clock speeds.
A die shrink from 90nm to 65nm is also on the cards along with a few more of those fancy stream processors for good measure. The GeForce 8800 GTX packs 128 of these shader processing units - expect the new chip to boast as many as 196.
For ATI, Nvidia's sworn enemy in the graphics business, the announcement is slightly sickening news. ATI's recently released Radeon HD 2900 XT is already well off the pace compared to the GeForce 8800. And that's despite being launched fully six months later.
If Nvidia really does launch a 1TFlop chip before the end of the year, things could get very ugly, very quickly for ATI.