In November, Nvidia reported that it plans to produce an ARM CPU for PCs in 2025. Now it appears that chips, or at least some versions, have been discovered to be calculated very slowly through Geekbench.
Geekbench’s chips are registered as N1X. This coincides with rumors from January that Nvidia will unleash a “high-end” N1X ARM-based chip this year, with more mainstream N1 variants continuing in 2026.
On the surface, the new Nvidia chip’s scores look terrible (via X user Jukan Los Reve). Earn 1,169 single core points on Geekbench 6. This is compared to Apple’s current M4 chip 3,831. This is also an arm base.
The multicore score is even weaker with just 2,417 points, 15,044 for the M4. And it’s a basic M4 chip. A MacBook with an M4 Max chip can violate 25,000 points or more than 10 times its performance.
In particular, the NVIDIA N1X list on the Geekbench results page shows a 4-core CPU. It’s almost certain that a “high-end” NVIDIA ARM CPU will have four or more cores. In fact, mainstream N1 variants are very likely to have more than four cores.
In other words, whatever appears on Geekbench rarely represents anything close to the final Nvidia N1X chip. This applies to the chip’s clock speed and firmware.
It is listed in Geekbench at 3.2 GHz. This is decent for the ARM CPU. But there’s no way to know if it’s actually running at that speed or if the chip is working perfectly in any other way.
The short story of the long story, what’s interesting here is the fact that, although it has nothing to do with the four cores and the 3.2 GHz specs, neither of them are final, the simple fact that a chip called the N1X runs somewhere and returns all sorts of benchmark results.
Of course, the idea of an arm-driven PC is not new. Qualcomm already has a serious tilt on the Snapdragon X-series chip, but Apple has proven that on M-series hardware, the ARM instruction set can go toe on x86 CPUs. In fact, when it comes to IPCs, Apple’s chips are currently a few miles away from Intel and AMD.
The problem for us at least, was always support for the game. Of almost all application types, games are the most challenging task for an entirely new set of CPU instructions.
Games on Snapdragon X via emulation have so far proven patchy. However, if any of the companies can make PC games work with ARM, it will be nvidia. It could potentially emulate a game developer who is very fast and convincing, like coding a native version of a game, or code a combination of both, via a very powerful CPU.
On the other hand, what will come to us with MediaTek bring about. One of the key drivers for MediaTek’s tie-in is believed to be support for 5G cellular networking, which may be important for the enterprise version of the new chip.
Consumer PCs definitely don’t need 5G. Frankly, I’d like to see Nvidia design CPU cores more than MediaTek. So we hope that MediaTek bits will actually be limited to wireless communications only.
For now, no other details are known. For example, which graphics architecture does the new APU use? All speculation for now, but the claim that N1X and N1 are on the N3 nodes of TSMC suggests that the chip can use NVIDIA’s next-generation Rubin GPU technology rather than the current Blackwell architecture.
This is because the GPU architecture is tied to some degree to handle the nodes. Blackwell is expected to be built on N4 and Rubin is expected to be at N3. At present, it is not impossible for Nvidia to port Blackwell to the N3 for its N1X chip. However, that requires a complete redesign of the new node, which takes time and money.
Perhaps it’s easier to use an architecture already built for TSMC’s N3 design rules. And it becomes Rubin.
Ultimately, you will know the time. But Nvidia’s true “high-end” arm tip with next-generation graphics is certainly exciting. I can’t wait to see what Nvidia is cooking.
