CES — Rumors have been swirling around the possibility of Nvidia developing its own x86-compatible microprocessor for some time now, given the moves by Intel and AMD to integrate graphics and chipset functions into their CPUs. However, from the top (CEO Jen-Hsun Huang) down, Nvidia consistently has had a single answer to any questions about such a possibility: "ARM is our CPU strategy," a statement we've long taken to refer to Nvidia's ARM-based Tegra SoC processors for phone and tablets. Today at its CES press conference, though, Nvidia finally pulled back the curtain its long-rumored CPU development project, and the news was stunning.
The firm has indeed been developing a CPU in secret, an effort known by the code-name Project Denver, and that CPU is intended to compete in the desktop PC and server segments. The words "high-performance computing" were even mentioned. But the processor in question will not be compatible with the dominant, Intel-owned x86 instruction set architecture. Instead, it will based on the ARM ISA that has, in the past several years, become almost entirely pervasive and incredibly widely used in mobile phones and tablets.
Huang described the Project Denver processor as a "fully custom" ARM-compatible CPU. Although Nvidia partnered with ARM and used ARM's ISA, we get the sense that the processor microarchiteture itself may be one of Nvidia's own making, with a higher end target in terms of performance (and presumably also power consumption) than existing ARM-compatible CPUs. We have almost no additional details on this CPU at present, but we can now see quite clearly the outlines of how Nvidia intends to use its chip design prowess to compete with Intel and AMD.
By virtue of its ARM compatibility, the Project Denver processor should be capable of running a range of popular operating systems out of the gate, including Google's Android and Apple's iOS. In addition, Huang hinted strongly that his news might be related to a possible announcement coming soon from Microsoft about a version of Windows tailored for ARM-compatible processors. Huang also pointed out that developers should be able to take advantage of the tools and development environments of the existing ARM ecosystems when developing applications for Nvidia's new CPU.
So ARM is indeed Nvidia's CPU strategy, but that strategy is not confined to pre-baked, licensed CPU cores aimed at low-power mobile devices. Instead, Nvidia intends to take ARM into new territories.
There's much we don't yet know about the Project Denver CPU, its projected performance, or when it might come to market. Nvidia's press release does mention that the CPU core will be "fully integrated on the same chip" as an Nvidia GPU, which makes sense. At least we now know what the rumors were about, and we're deeply intrigued to see how this story plays out from here.
|Biostar's Ryzen motherboards race toward release||52|
|TSUBAME3.0 gears up for AI supercomputing with 2160 Tesla P100s||26|
|Master of Shapes brings Vive tracking to Daydream VR||4|
|Deals of the week: Z270 motherboards, storage, and more||15|
|Phanteks Glacier gear flows into the water-cooling market||11|
|Display your graphics card with Thermaltake's PCIe riser cable||23|
|WWDC 2017 returns to its roots in San Jose||3|
|Unreal Engine 4.15 arrives with HDR and AFR support||60|
|MSI Aero ITX graphics cards put Pascal in petite places||5|