Nvidia To Offer GPUs Integrated Onto ARM-Based Processors

The integration of graphics capability and central processing on a single chip continues, with GPU specialist Nvidia and British company ARM planning to bring their chip technologies together.

Nvidia on Thursday at CES said its upcoming high-performance processors, code-named Project Denver, will combine Nvidia's GPU technology with a CPU reference architecture licensed from ARM. According to Nvidia, the result will be a heterogeneous platform offering performance as well as efficiency on devices ranging from laptops to data centers.

"Since last year, our CPU strategy has been ARM, and we're very excited about that," an Nvidia spokesperson told CRN. "We spent the past three years developing our CPU strategy, and it's now fully focused on ARM."

The spokesperson also said Nvidia will be integrating the ARM-based CPU onto its high-powered Maxwell GPU architecture, which the spokesperson said Nvidia expects to bring to market in the next several years.

An Nvidia blog post about Project Denver appeared on Tuesday saying the company plans to leverage the ARM-based CPUs for products in a wide range of markets including servers, desktops, and even supercomputers.

In the post, Nvidia says ARM's architecture and open business model have made it the standard architecture for mobile PCs and other embedded devices. ARM claims the largest share of the mobile processor market. In the third quarter of 2010, ARM’s customers reported about a 35 percent increase in shipments of mobile handset chips compared to the third quarter of 2009. ARM earned 61 percent share of the mobile market in unit shipments in the third quarter of 2010.

ARM declined to discuss the architecture it will license to Nvidia. Nvidia currently licenses the Cortex-A9 architecture for its current Tegra 2 chips.

Nor is Project Denver the first instance of an ARM-based processor powering servers. In May, ARM announced it was running one of its Web sites on its own processors, despite the fact that the company's business model is predicated on licensing its reference designs to other manufacturers. In November chipmaker Marvell launched its Armada XP quad-core ARM-based server processors.

The breakthrough that allowed this latest partnership between semiconductor companies actually involves new support from the operating system.

Next: Competition Mounts Over Integrated Graphics Solutions

Microsoft has announced that it is bringing Windows to ARM-based CPUs provides, allowing Nvidia to leverage ARM-based PCs based for the Denver Project.

"They showed the Windows platform running on ARM at CES," Nvidia's spokesperson said. "That was very exciting for Nvidia."

The spokesperson added that Nvidia is not announcing product details at this point.

Nevertheless, high-performance GPU computing is central to Nvidia's product strategy. Nvidia's CUDA parallel processing platform for general-purpose GPU computing is expanding into a variety of applications. Nvidia CEO and co-founder Jen-Hsun Huang, speaking at Nvidia's GTC conference in September, said that nine OEMs are now using Nvidia's supercomputer chip architecture, code-named Tesla, compared to just one last year. According to a report issued in November, Nvidia's GPUs offer supercomputers accelerated computation on three of the world's top five most powerful systems.

Meanwhile at CES, rival chipmaker AMD has launched its Fusion processors, which integrate multicore CPU technology with HD video acceleration and high-performance GPU processing on a single die.

In addition, the most anticipated integrated graphics solution, Intel's Sandy Bridge, launched on the first day of CES, with CEO Paul Otellini calling it the best product Intel has ever built. Sandy Bridge features built-in video and graphics capability for updated versions of Intel's Core i3, i5 and i7 series of processors, which Intel originally introduced at last year's CES.

Whether it ends up being the best offering in the growing integrated graphics segment remains to be seen.