Nvidia announced today that its next-generation graphics chips will be increasingly power-efficient in an age when power requirements are limiting what computers can do.
Jen-Hsun Huang, chief executive of the Santa Clara, Calif.-based graphics chip maker, said at the company’s GPU Technology conference today that Nvidia will introduce new graphics chips in 2011 and 2013, the first code-named Kepler, the second Maxwell.
The Kepler chip, slated for 2011, will be three to four times more power efficient that the company’s mainstay Fermi chip, which was introduced in March of this year. Maxwell, which arrives in 2013, will be 16 times more efficient than Fermi. In a press conference after his keynote speech, Huang said that Nvidia would shift to introducing big families of new chips every two years, instead of doing that once every year as it had done in past years. Huang said the reason for that is that Nvidia has to rely on manufacturing upgrades for chip factories that come along every couple of years. [Nvidia points out that it has consistently introduced major architectural changes every couple of years, with some exceptions when it runs into delays].
But Huang said that Nvidia will still redesign its chips and introduce new ones just about every year. These “mid-life” kickers will be redesigns of chips that use a current manufacturing process. Nvidia designs graphics chips and then hands them over to contract manufacturer Taiwan Semiconductor Manufacturing Co., which fabricates them. Nvidia then sells the chips to computer makers and add-on board makers. TSMC upgrades its chip factories to a new manufacturing process technology every couple of years.
Huang noted that Nvidia doesn’t talk about its roadmap. Rivals will note that the chips talked about today are “vaporware.” In fact, in October of last year, Huang stood up at the company’s conference holding up a Fermi-based graphics card. But Nvidia didn’t introduce Fermi-based chips until March. But Huang said there are hundreds of engineers working on the Kepler and the Maxwell chips.
Nvidia didn’t reveal anything new about its Tegra chips for cell phones and tablet computers. In the press event, Huang said the Tegra business is a year behind his goals, but he noted the adoption of Tegra 2 (which is a microprocessor with graphics that uses an ARM-based processor) is “phenomenal.” He said that Tegra 2 chips will be used in a class of cell phones known as “superphones,” the topic of VentureBeat’s MobileBeat conference in July.
“These superphones are going to be really important to us, and the tablets too,” Huang said.