The newest rumors about the specifications of Nvidia’s anticipated first PC CPU, code-named X, indicate that the company will use off-the-shelf ARM Cortex X5 cores for the processor portion of its 2025 device, along with brand-new Blackwell GPU cores. The more peculiar rumor, meanwhile, suggests that Nvidia plans to use Intel’s fabrication facilities to produce yet another silicon nail in the x86 coffin.
After all the hype this year about AI PCs, Nvidia’s decision to develop an ARM-based processor makes perfect sense, especially in light of earlier reports of the green team’s collaboration with MediaTek to provide laptop reference designs. It would be advantageous for the company if it could produce laptops that are exclusively based on its own CPUs and graphics cards. This would be especially true if it could expand into the lower-end market, where discrete GPUs are not typically crammed into thin and light laptops.
Nvidia therefore creates a system on chip (SoC) that integrates its own graphics silicon as an integrated GPU for a low-power/high-performance chip that it can subsequently sell to other notebook manufacturers, using Arm’s CPU cores as a basis. It seems obvious to do that.
According to a rumor on specs, it will have a Blackwell iGPU, a Cortex X5 CPU cluster, and LPDDR6 memory. It comes from @XpeaGPU on X. Additionally, it says that TSMC’s N3P node will be used in its construction. However, another X rumor-monger, @Kepler_L2, swiftly refuted this assertion. They assert that Nvidia will actually employ an Intel 3nm manufacturing technique.
They might just be referring to Intel 3, as Intel doesn’t actually have a 3nm node. However, it might also refer to the Intel 3-T upgrade, which has the ability to stack chips or wafers from many foundries together using Fosteros Direct 3D (PDF warning). In other words, Nvidia might continue to assemble its Blackwell GPU cores using TSMC N3P, while Intel would be responsible for producing the ARM Cortex X5 cores and the foundation tile that houses them all.
Intel Foundry Process Node Roadmap
For some time now, there have been rumors that Nvidia may consider Intel as a different foundry partner to TSMC. Some people even speculate that Intel may produce different GPU lines. This would have required additional labor to port designs from one production facility to another, which would have been odd given the various packaging techniques used by the two foundries. However, that becomes much more likely if Intel is only being used to develop the processing part and possibly the interposer.
But none of this would indicate that Nvidia would get bespoke cores, even if Apple and Qualcomm have already adopted that strategy for their own new chips. However, that might not be entirely detrimental to the green team. The Nvidia Drive platform in the automotive space uses a similar straightforward process to pick pre-made cores from Arm’s catalogue and avoid having to mess with architecting your own processor cores. However, doing so may limit your ability to precisely control power levels or add in specific units to help deal with x86 emulation layers.
Rosetta’s remarkable efficacy in translating x86 software for Apple’s M-series chips might be attributed, at least in part, to this feature, which the company proudly highlighted as being unique to its own bespoke chips. Given that Microsoft has stated that Prism will be just as beneficial for Windows as Rosetta was for iOS, it is anticipated that Qualcomm has perhaps done something akin with its Oryon cores for the Snapdragon X Elite CPUs.
Given that the ARM Cortex X5 CPU cores are codenamed Blackhawk, Nvidia’s marriage of the Blackwell GPU architecture with them also has a charming symmetry.
When will Nvidia’s own CPUs be available for purchase? Michael Dell, the founder of Dell Technologies and Mr. Not-just-a-clever-name, says we should “come back next year” to evaluate where the company is in the AI PC market. In a Bloomberg interview, he sat next to Jen-Hsun Huang and delivered the joke.
We’re not sure, though, if this implies Nvidia laptops would control the entire hardware ecosystem, similar to Apple laptops. It’s possible that the green team is merely preparing SoCs for other laptop makers to enable them to eventually incorporate them into the thin-and-light market without discrete GPUs, but I doubt Nvidia could resist producing a reference design that is at least reminiscent of the MacBook.