Nvidia CEO: Tegra K1 ARM Chip May Be Used in Microservers
Nvidia's Tegra K1, the powerful mobile chip introduced at the 2014 Consumer Electronics Show in January, could make its way into low-power dense servers, according to the company's CEO.
- Navigating Your IT Career
- Exploring the Private Cloud for Your Organization
- IT Manager's Guide to Social Networking
During a conference call with analysts and journalists to discuss Nvidia's first-quarter financial numbers, CEO Jen-Hsun Huang said that while the 192-GPU core processor—which will come in 32- and 64-bit versions—is being aimed at products in the automotive and gaming spaces, as well as for differentiated products, there also has been interest from data center system vendors.
"I think we're seeing a lot of interest in putting something like Tegra in microservers, but one step at a time," Huang said May 9.
Microservers are highly dense, low-power systems designed to run large numbers of small workloads, the kinds of which are found in hyperscale data centers run by the likes of Google, Facebook and Amazon. It's a point of competition between Intel and ARM, which sees a market for the highly energy-efficient systems-on-a-chip (SoC) designs that are found in most smartphones and tablets. ARM has taken a major step forward in pushing its data center ambitions over the past several months as it licenses the 64-bit ARMv8-A architecture to a range of chip makers, including Advanced Micro Devices, Applied Micro, Samsung and Marvell Technologies.
AMD, which this month unveiled an expanded road map for its ARM-based efforts, this year will ship its Opteron A1100 "Seattle" ARM-based server chip.
Intel, which dominates the server space with its x86 architecture, is pushing its low-power Atom platform for the energy-efficient microserver market. The chip maker last year rolled out its second-generation chip for the space, the C2000 "Avoton," and later this year is due to roll out the 14-nanometer "Denverton" SoC.
Nvidia's Tegra K1—which succeeded the Tegra 4 chip—includes a 192-core Keplar GPU and either a 32-bit quad-core CPU or, for the 64-bit version, the ARMv8-A architecture. Huang said he expects the 64-bit Tegra K1 "to be in production long before the end of the year and be the first high performance 64-bit processor in the market."
A key point for the chip will be the software stack. The company is continuing to add software to its Grid GPU stack, and that software "can very well eventually be used on top of Tegra as well."
Read more on "Server Hardware Spotlight" »