Connecting the CPU and GPU—Intel and PCIe

In early 2009 Intel sued Nvidia over their 2004 chipset licensing agreement that let Nvidia make core-logic (chipsets) for Intel in exchange for Intel licensing NVIDIA's 3D, GPU and other patents. Nvidia made the chips for their nForce series chipset and the two-chip ION Platform.   ION's combination of Nvidia and Intel chipsets provided a 10x performance advantage over Intel-only chipsets and was popular in Apple laptop computers, etc. Intel claimed their agreement did not cover their newer Nehalem architecture, which featured an integrated memory controller. Nvidia chose to countersue. The USD 1.5 billion settlement in January 2011 barred Nvidia from making CPUs with Intel's x86 technology. Nvidia GPUs have since been relegated to connecting to Intel CPUs via the relatively slow PCIe buses. This is also why, at present, IBM's Power CPUs are connected to Nvidia's GPUs over NVLINK, while no such connections exist for Intel CPUs. AMD have similarly connected their CPUs and GPUs via their HyperTransport technology.