The antitrust suit brought by the Federal Trade Commission against Intel last December may be at an end, with the parties having reached a tentative settlement agreement.
The suit against Intel alleged that the company had illegally stifled competition in the processor market. Specifically, it said that the chip maker threatened PC makers (including Dell, Hewlett-Packard, and IBM) who wanted to source processors from AMD, then rewarded them for remaining loyal to Intel. The settlement prohibits Intel from making such threats and offering rewards in return for exclusivity—the company can no longer punish PC makers for buying processors from AMD.
The settlement also imposes four specific restrictions on Intel in order to ensure that markets remain competitive. First, Intel must alter its intellectual property licensing agreements with AMD, NVIDIA, and Via, to allow those companies greater freedom to collaborate or merge without risking a suit from Intel for patent infringement. This should prevent replays of the squabble between AMD and Intel when AMD divested itself of its manufacturing facilities, a move arguably in breach of the licensing terms between the two companies.
Second, Intel must extend its x86 license agreement with Via for another five years beyond the expiration of its current agreement, which is set to expire in 2013.
Third, Intel must retain the use of the PCI Express interface for at least six years, such that GPUs can be used without suffering any performance limit. The intent is to ensure that companies like NVIDIA still have a reason to produce GPUs targeted at Intel-based systems, even when those Intel systems contain integrated GPUs.
Finally, Intel must inform software developers that its compiler discriminates between Intel and non-Intel CPUs, and that it produces code that may not use all the features of those non-Intel CPUs. The company will also have to reimburse any developer who wants to compile their software using a non-Intel compiler.
Preventing total market dominance
The settlement is designed to ensure not just that the CPU market remains competitive, but also that the GPU market continues to remain that way. Intel is already the dominant maker of GPUs, courtesy of its widespread integrated GPUs. With the latest generation of processors including a GPU in the CPU package, and the next generation set to include it on the CPU die, this domination is likely to continue.
The growth of integrated GPUs poses a particular problem for NVIDIA. The GPU company has no license for the DMI and QPI buses used on Intel's current processors, which prevents it from being able to make replacement chipsets. This is one of the reasons why Apple stuck with an older, FSB-based processor for the 13" MacBook Pro. NVIDIA's only option for the new processors is to produce discrete PCI Express devices.
This is itself an imperfect solution; the PCI Express bus makes low-power, low-cost GPUs that share system memory impractical, which is detrimental to NVIDIA's Ion platform. But at least PCI Express does give the company some options.
Intel might have been tempted to ditch PCI Express completely, at least on some of its processors and chipsets; for many Atom machines, the combination of USB and SATA is good enough, making PCI Express superfluous. The settlement agreement means that the company will have to retain the interface—at least for the next six years—ensuring that NVIDIA always has some ability to provide GPUs for Intel systems, even if it is not as extensive as NVIDIA might like.
The FTC said that, though the new agreement goes beyond the terms and restrictions imposed after previous Intel suits, it will nonetheless leave Intel room to continue to innovate and compete. The settlement will be subject to public comment for 30 days, at which point the Commission will decide whether to make it final.
Source: ars technica
