Nvidia talks up Fermi's replacement, Kepler

Nvidia surprised attendees at its GTC 2010 conference with a sneak preview of its next-generation replacement for Fermi, codenamed Kepler.

During his keynote speech, Nvidia's Jen-Hsung Huang announced that Fermi's replacement will launch "in late 2011" and offer three to four times the performance of the company's current high-end chip.

For the nay-sayers in the red camp, who point to the comically high power demands of early Fermi models in derision, Huang was pleased to announce that the performance boost comes at little to no cost in terms of increased power consumption, thanks to a team of "hundreds of engineers" working to avoid a repeat of the marketing nightmare that was the Fermi launch.

Sadly, Huang was less forthcoming with firm details about exactly what Kepler will actually be: we know it's the next-gen GPU replacement for Fermi, and we know that it will offer significantly improved performance over its predecessor - but what we don't know is how much it will cost, what its actual specifications are, or exactly when it will launch.

Peering even deeper into his crystal ball, and offering even fewer firm and confirmed details, Huang went as far as to detail the replacement for the replacement - Maxwell.

Designed to replace Kepler in 2013, Maxwell will offer a boost of around twelve times the performance of current-generation Fermi chips - or around three times again the performance of Kepler.

With a growing increase in demand for multi-monitor gaming, and Nvidia trying to push its 3DVision immersive gaming technology - plus the growth in adoption of Nvidia's CUDA general-purpose GPU processing technology in the mini-supercomputer sector - the news of a new and more powerful CPU range is welcome, but Nvidia really needs to look to an earlier release for Kepler if it wants to steal a march on rival AMD.