SAN JOSE, Calif. – ARM will triumph over Intel as smartphones and tablets disrupt the x86 PC industry, said Jen-Hsun Huang, chief executive of Nvidia, speaking at the company's annual conference.
"It's a foregone conclusion that the personal computer of the future is this size," said Huang, holding up his smartphone. "You could add wireless HDMI to it someday, and it could also be your set-top box," he said in a fireside chat at the event.
"The PC of the future will be made by new OEMs, sold through new distributors and use a new instruction-set architecture," Huang said. "ARM will be the most important CPU architecture of the future, and it already is the fastest growing processor architecture," he added.
Nvidia has been selling an ARM-based chip called Tegra for mobile consumer devices. It has won design wins in the Microsoft Zune, but not appeared in any tablet or tier-one smartphones to date.
Huang would not comment on whether any Tegra-based tablets will emerge before the end of the year. However, he did say "Motorola, Samsung and LG are very important to us and will represent long term some of our biggest customers if we are successful."
Asked about Nvidia's relationship with x86 giant Intel, Huang quipped, "we're trying to have a lesser one."
The two companies have been locked in legal disputes over technology licensing.
Both Intel and Advanced Micro Devices will launch families of processors next year using x86 cores and their own internally designed graphics cores. Some analysts say those chips will wipe out the market for low-end discrete graphics chips, but a market for high-end discrete graphics chips is expected to remain robust.
With its core PC business threatened by the upcoming hybrid chips, Nvidia is trying to develop markets for its Tegra chips in mobile systems. In addition, it is using its conference primarily to develop so-called GPU computing markets for its Tesla processors across a wide variety of high performance vertical markets.
Nvidia rolled out to mixed reviews its first Tesla chip, called Fermi, at its inaugural conference last year. This year it said it will release in the second half of 2011 a next generation, called Kepler, made in a 28nm process and is working on a follow-on chip called Maxwell. The company provided few details on those chips.
"This is the year when apps delivered for GPU computing go to production," said Huang in his talk.
The market is accelerating the adoption to parallel programming tools such as Nvidia's CUDA environment, Huang said.
"We are doing n more fundamental computer science than we have done in a long time," Huang said. "I don’t remember in the past 30 years this much change being about to happen," he added.
CEO Jen-Hsun Huang has driven Nvidia into mobile and vertical markets as Intel and AMD prepare processors using x86 and graphics cores.
"They have picked the right strategy" with GPU computing, said Jon Peddie, principal of market watcher Jon Peddie Research (Tiburon, Calif.).
The GPU computing market is still small but could grow exponentially, Peddie said. He estimated the industry will ship in 2010 about 150,000 of the high-end graphics chips used in specialty computing markets, well below the estimated 12 million graphics chips a quarter Nvidia alone now ships to PCs.
However, Peddie cautioned that it's hard even for companies such as Nvidia to track which of its chips wind up in high-end graphics computing applications.