Nvidia, Intel May Settle Case, But Chipsets Aren't Coming Back
The original dispute between the two companies was over whether or not Nvidia had the right to design chipsets for Intel products with an integrated memory controller. Later, the issue expanded to cover Intel's Atom platforms and its decision to sell complete Atom SoCs for less than the price of an Atom processor. The FTC recently dealt with this second item, ruling that Intel must include PCIe support for at least the next six years in all its products, but was not required to redesign its upcoming Oak Trail Atom products to support PCIe.
According to various little birds, Intel will soon pay Nvidia $1 billion in exchange for being darn mean and will receive access to certain NV patents (probably related to SLI) in exchange. No discussion of NV and Intel would be complete without mentioning an x86 license, but as always,
Intel is supposedly refusing to grant a request that NV might or might not have made in the past, currently, or will make at some point in the future. Possession of an x86 license does not magically hand the new owner a competitive x86 processor; it would take NV at least 18-24 months to design a new architecture, even if it had a CPU team fully formed and waiting right this minute.
Nvidia's core graphics business is doing relatively well; later iterations of Fermi are proving more competitive in terms of heat/power than the original GF100 products.
Recent comments by Jsen-Hsun underscore the point that NV won't return to the chipset business. At the company's earnings call last week, Huang stated that the company is now focused on building Tegra SoC's and is allowing the chipset side of things to slowly wind down. When asked about its AMD business, Huang's comments were a bit surprising. "On the AMD side, our AMD chipset remains quite well positioned. My sense is that our chipset there will continue to ship throughout next year."
"Well-positioned" is not how we'd describe the nForce 980a chipset, unless it includes the phrase "to be kicked out the door." NewEgg lists 18 best-selling AMD boards; 17 of which are AMD chipset solutions. Similarly, there are 141 AMD motherboards listed in NewEgg's database and just six motherboards total (leaving out the Open Box models) from three generations of Nvidia's chipsets.
What we'd really like to see is some actual product launches to build momentum around Tegra 2. Nvidia's chipset business was never really the company's core competence; it was never clear how NV would differentiate its own chipsets from Intel's once GPU's moved on-die. About two months ago we wrote an article discussing Tegra 2 and the platform's slow path to market. We haven't seen much change since that article was written, but even if Tegra 2 devices slip past Christmas, it's important that Nvidia be able to demonstrate real products with concrete release dates come CES.
Last year, Tegra 2 was one of the major features of the show—it'd be a shame for it to star as a farce of itself the second time around.