A groundbreaking collaboration between Intel and Nvidia has been announced, signaling a new era for computing hardware. The two industry giants plan to merge Intel's x86 processors with Nvidia's graphics processing units to develop innovative products for both consumer PCs and AI servers. This strategic alliance is further solidified by Nvidia's substantial $5 billion investment in Intel. The specifics of these forthcoming chips remain largely under wraps, though an existing Nvidia product, the GB10 'Superchip,' offers intriguing clues about the potential design and integration strategies.

The GB10 'Superchip,' a result of Nvidia's partnership with Mediatek, already demonstrates a similar collaborative approach. While it incorporates an Arm CPU rather than an x86 architecture, its design provides a compelling template. A key feature of the GB10 is its use of Nvidia's high-bandwidth, low-latency NVLink interface to connect the CPU and GPU dies. This technology is explicitly mentioned as a critical component of the new Intel-Nvidia venture, particularly for consumer PCs where custom x86 System-on-Chips (SOCs) will integrate Nvidia RTX GPU chiplets.

In the GB10 architecture, Mediatek provides the SoC die, encompassing CPU cores, memory controllers, and external interconnects like USB and PCIe. An Nvidia GPU die is then integrated into the same package, with the crucial connection facilitated by an NVLink-based 'C2C' (chip-to-chip) interface. This method significantly enhances communication speed and reduces latency compared to traditional PCI Express connections between CPU and GPU. The success of this design in the GB10 strongly suggests that the Intel-Nvidia collaboration will adopt a similar NVLink-centric approach, aiming for superior performance through tighter integration.

The possibility exists that Nvidia might reuse the GPU die from the GB10 in the initial generation of Intel-Nvidia chips. This could accelerate development and reduce costs, allowing for a quicker market entry. However, this scenario relies on Intel developing an NVLink-compatible CPU SoC, which could take considerable time. If this proves viable, even an existing GB10 GPU die, offering performance comparable to a desktop RTX 5070 graphics card, would deliver excellent graphics capabilities for several years, making such a move strategically sound for Nvidia.

Despite the potential benefits of leveraging the GB10's architecture, there are cost implications to consider. For instance, Nvidia's DGX Spark machine, which utilizes the GB10 chip, carries a price tag of $4,000. While it's improbable that consumer PCs resulting from this collaboration would reach such high prices, the cost of advanced integration technology could mean that these new chips might not immediately lead to inexpensive laptops with exceptional gaming performance. The full details of these innovative chips, including their specifications and market positioning, will likely unfold over time, with the GB10 offering a valuable sneak peek into their possible form.