Among other tidbits that emerged from the recent Intel-Nvidia love in is confirmation of what we'd long suspected, namely that the upcoming Nvidia N1 Arm chip for PCs is indeed the very same as GB10, the "Superchip" at the heart of the DGX Spark AI box.
"We also have a new Arm product that's called N1," said Nvidia CEO Jensen Huang at the press conference for the newly announced Intel-Nvidia alliance. "That processor is going to go into the DGX Spark and many other versions of products like that. And so we're super excited about the Arm road map, and this doesn't affect any of that."
Taken at face value, that's both good and bad news. For starters, it means we have a good idea of the specifications of the chip, which Nvidia has already published, if not quite such a clear feel for its real-world performance. That's especially true in the context of PC gaming as opposed to the AI workloads the DGX Spark is aimed at.
Specifically, that means an Arm CPU SoC with 10 Arm Cortex X925 cores and 10 Arm Cortex A725 cores. To that is attached an Nvidia GPU chiplet with 6,144 CUDA cores.
On paper, that gives N1 exactly the same CUDA core count as the RTX 5070, albeit the RTX 5070 runs with a partially disabled GB205 chip, the latter containing 6,400 CUDA cores all told. Of course, it's far from clear what configuration Nvidia might use for N1 when targeted at PCs.
But arguably a bigger question mark hangs over the broader issue of playing PC games on Arm CPUs. Despite the arrival of the Qualcomm Snapdragon X series of Arm CPUs for PCs and a push from Microsoft to support Arm better on Windows, in part courtesy of the Prism translation layer for running x86 code on Arm CPU cores, PC game support and performance on Arm chips is patchy at best.
Now, if any company has the resources and will power to get games running nicely on an Arm CPU as opposed to an x86 CPU, you would think that company is Nvidia. It has relationships with most if not all major game developers and endless cash to spend, should it choose.
What's more, emulation needn't be a dirty word when it comes to PC gaming. Apple has shown how effective emulation can be when done right with its Apple silicon chips offering very high performance x86 emulation.
Of course, Apple added specific hardware features to its custom Arm cores to support that, while the Nvidia N1 chip uses off-the-shelf CPU cores designed by Arm itself. But emulation can definitely work.
Indeed, the Proton compatibility layer for Valve's SteamOS, as used by the Steam Deck proves that translating code on the fly can be a goer, even if Proton is a slightly different beast in translating between Microsoft's DirectX API and Vulkan, while still running on x86 CPUs in both cases.
But the basic question remains: Regardless of how much theoretical GPU power N1 offers, will its Arm CPU cores play nicely with games?
Beyond all that there's the question of price. The DGX Spark is a $4,000 box, which is pretty scary in a PC context. But then the DGX Spark is a pretty specific kind of machine aimed at AI worklaods and comes packed with 128 GB of RAM, Nvidia's ConnectX network interface and more. It's not trying to be built down to the price of an RTX 5070 gaming PC.
On the subject of the price, the obvious comparison here is AMD's Strix Halo chip, another chiplet-based APU with a big fat GPU. That too is arguably too expensive to be interesting as a PC gaming solution and, a bit like DGX Spark, is basically a chip for running AI models locally.
Oh, yes, and the DGX Spark has missed its original launch window of July and still hasn't shipped to customers at the time of writing, amid rumours that the whole Nvidia Arm CPU for PCs thing is being delayed until at least early next year.
All of which means we're a little cool on N1. It has plenty to prove, that's for sure.
