Over the past few months the once frosty relationship between Intel and Nvidia has palpably thawed. Anecdotally the roots of the animosity between the two companies hark back to an article that was published in Wired magazine in 2002 in which Nvidia’s VP of Investor Relations commented that ‘In 10 years, we should be bigger than Intel’.
In the intervening years Intel made many moves to block Nvidia, the most important of which being the effective killing of Nvidia’s chipset division (at one point the main focus of the company) with the launch of Intel’s Nehalem architecture. Nehalem moved the all-important memory controller onto the CPU and off the chipset, and Intel’s line was that Nvidia only had permission to make chipsets for processors that didn’t have a memory controller.
This issue has been simmering in the courts ever since, or at least it had up until late last year, where both companies announced (seemingly out of the blue) that they were delaying an impending court date. Within a week we started to see press releases appearing in the leadup to the launch of Intel’s Sandy Bridge processors that actively encouraged the pairing of Intel CPU and Nvidia GPU, a logical move considering that the companies share a mutual competitor in the shape of AMD.
While Sandy Bridge definitely raises the historically low bar set by Intel for 3D graphics performance, its graphics processor lacks key high end functionality like support for DirectX 11 and GPGPU computing. Serious and not-so-serious gamers will still likely be looking for a discreet graphics card, so Intel needed Nvidia’s support lest it push its customers to buy AMD graphics to pair with the Intel CPU.
This move beyond the animosity of the past decade makes sense, but hours ago both companies made an announcement that very few would have predicted a few months ago. Not only have Intel and Nvidia settled their differences, they have announced a cross licensing deal in which Intel will pay Nvidia $US1.5 billion over the next five years. It will give Intel access to Nvidia’s all important graphics patents, and gives Nvidia partial access to Intel’s patents.
Key to the historical animosity is that the agreement does not allow Nvidia to build an x86 compatible CPU (perhaps the top Nvidia rumour of the past five or so years). It also doesn’t allow Nvidia to build chipsets for Intel CPUs with integrated memory controllers – effectively locking Nvidia out of the x86 chipset market permanently. One suspects that the main benefit of access to Intel patents will actually come in the form of the recently announced ‘Project Denver’ in which Nvidia and ARM are to collaborate to bring the ARM architecture to desktop and servers.
As for Intel, well it finally gets access to Nvidia’s patents. This should mean good things for consumers, allowing Intel to further improve upon the graphics parts of its CPU lineup. It also means that it has more of a chance to compete against AMD’s Fusion APUs (Accelerated Processing Units), which combine x86 CPU cores with high end 3D graphics. Of course, given the long development times inherent in seminconductor manufacture, we don’t expect to see Nvidia features popping up in Intel processors just yet, but what today’s announcement does do is accelerate the rate at which Intel’s graphics will improve.
Which is a win for pretty much everybody.