For the last 20 years, NVIDIA has developed and used its own graphics IP in-house. While the company engaged in several prominent cross-licensing deals with companies like Intel, those deals never resulted in an Intel-branded graphics card or an equivalent product from another company. Today, that changes -- NVIDIA has announced that henceforth, it will license its graphics designs to other companies.
In a blog post today, company spokesperson David Shannon writes: "our next step is to license our GPU cores and visual computing patent portfolio to device manufacturers to serve the needs of a large piece of the market... We’ll start by licensing the GPU core based on the NVIDIA Kepler architecture, the world’s most advanced, most efficient GPU. Its DX11, OpenGL 4.3, and GPGPU capabilities, along with vastly superior performance and efficiency, create a new class of licensable GPU cores. Through our efforts designing Tegra into mobile devices, we’ve gained valuable experience designing for the smallest power envelopes. As a result, Kepler can operate in a half-watt power envelope, making it scalable from smartphones to supercomputers."
And Now For Something Completely Different
NVIDIA tries to spin this as an approach it's taken before with, for example, the PS3. That's a bad example, in our opinion. The PS3'
s Reality Synthesizer (RSX) was a custom core based on NVIDIA hardware that the company developed in partnership with Sony. While it's true that Sony had a license to build an NVIDIA design, that's a bad lens for evaluating this decision. By opening its Kepler architecture up for licensing, NVIDIA is pursuing an ARM-like strategy of focusing on core IP while providing a basic design that other companies can license.
It's not hard to see why. Imagination Technologies is currently the graphics darling of the mobile industry. It has provided GPU solutions to Apple, Intel, Samsung, and a host of other, smaller players. With Tegra, NVIDIA tackled the market with an all-in-one solution that it hoped would win designs on the strength of its overall performance and mobile graphics. Clearly, NVIDIA now feels it has more to offer beyond just a silicon solution, by licensing its graphics IP separately, rather than trying to entice manufacturers with only packaged silicon.
That said, it's not clear who wants to license NVIDIA's IP right now. The company has announced no takers at this junction. Samsung, as the reigning "I'll try anything once" champion of the mobile market is a logical suspect. Qualcomm
is unlikely, given Adreno's strength. Intel is working on scaling its own graphics technology down to smaller form factors; the upcoming 22nm Bay Trail hardware will use Intel's own graphics. Texas Instruments' mobile division is dead.
is the only top-tier manufacturer that might be interested in NVIDIA's IP as opposed to sticking with Imagination Technologies, but there are a host of smaller vendors, including Broadcom
, and Lenovo
. It's also possible that this is a move aimed at grabbing market share in Asian or Indian markets, where the competitive landscape is very different.
There's one more significant difference between NVIDIA and ARM that we want to highlight -- ARM doesn't compete with its own customers. NVIDIA, presumably, would. That's not a trivial distinction, particularly in markets where every tenth of a watt matters and performance is sharply constrained by thermals. We'll have to wait and see how this plays out. At its heart, though, this is an admission that NVIDIA isn't going to control the future of computing via Tegra, any more than it could stop the march of CPU-GPU integration by building better GeForce
cards. Recognizing that now, and licensing out GeForce IP, is the best way for NVIDIA to ensure its technology remains part of the mobile ecosystem down the road.