Nvidia's long-awaited N1/N1X SoC specs leak ahead of Computex launch — N1 to feature up to 20 Arm-based cores, standard N1 equipped with 12- and 10-core configs
The top-end N1X SKU is essentially a rebranded GB10.
It's finally happening — Nvidia is set to launch the N1 family of SoCs at Computex tomorrow after years of it being stuck in the rumor mill. And just a day before the grand reveal, most of the specs for the lineup have leaked, courtesy of Videocardz. As expected, there seem to be two distinct SKUs: N1 and N1X, both targeting different performance and price levels. Take the following information with a grain of salt since it comes from documents dating back to 2024.
We begin with the standard N1 that reportedly comes in two configs. There's a 12-core (8+4) model with 2,560 CUDA cores and a 10-core (7+3) model with 2,048 CUDA cores. Either variant comes with 8x PCIe 5.0 lanes and 3x PCIe 4.0 lanes; the base N1 supports up to two M.2 SSDs. Memory support tops out at 64GB across 8x LPDDR5X channels. In terms of power budget, we're looking at 18W-45W TDPs.
Then there's the more exciting N1X that, at the top-end, is identical to the GB10 found in Nvidia's DGX Spark mini-PC, which Nvidia CEO Jensen Huang already confirmed. There are two SKUs for this SoC as well: a full-fat 20-core (10+10) monster with 6,144 CUDA Cores — same as the desktop RTX 5070 — and an 18-core (9+9) variant with 5,120 CUDA Cores. Both chips have the same power envelope of 45W-80W.
N1X will reportedly start at 16GB LPPDR5X configs, but the platform can support up to 128GB across 16 channels. We don't know at what speeds exactly yet, but a previous leak suggested these chips are operating at 8,533 MT/s, which would make the RAM config faster than AMD's Strix Halo. The N1X also has 12x PCIe 5.0 lanes and 5x PCIe 4.0 lanes capable of accepting up to three M.2 SSDs.
Model | CPU Cores | GPU (Cuda Cores) | PCIe Lanes | Memory (LPDDR5X) | TDP |
|---|---|---|---|---|---|
Nvidia N1X (1) | 20 (10+10) | 6,144 | 12x PCIe 5.0 + 5x PCIe 4.0 | 16GB to 128GB - 16 channels | 45W-80W |
Nvidia N1X (2) | 18 (9+9) | 5,120 | 12x PCIe 5.0 + 5x PCIe 4.0 | 16GB to 128GB - 16 channels | 45W-80W |
Nvidia N1 (1) | 12 (8+4) | 2,560 | 8x PCIe 5.0 + 3x PCIe 4.0 | 8GB to 64GB - 8 channels | 18W-45W |
Nvidia N1 (2) | 10 (7+3) | 2,048 | 8x PCIe 5.0 + 3x PCIe 4.0 | 8GB to 64GB - 8 channels | 18W-45W |
All signs point toward Nvidia unveiling the N1/N1X family tomorrow, so we should know all about these chips very soon. Based on the history of leaks and rumors, and educated guesses, these specs make sense. Once made official, Nvidia would re-enter the laptop market, directly competing with AMD, Intel, and, of course, Apple. We say re-enter because the company already tried selling ARM-based PC chips back in 2011.
The advent of a new mainstream SoC from Nvidia could open doors to exciting handhelds, OEM PCs, and perhaps even a refreshed Shield TV that fans have been yearning for forever. It's a positive development considering just how abandoned consumers and, in particular, gamers have felt these past few years, but it all comes down to pricing. And during a RAM crisis, that's not going to be ideal.
N1X will likely target the $2,000+ market, competing with the MacBook Pro, but the N1 could be an exciting midrange option under $1,500.
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he’s not working, you’ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun.
-
bit_user Reply
Gorgon Halo uses that same speed:The article said:We don't know at what speeds exactly yet, but a previous leak suggested these chips are operating at 8,533 MT/s, which would make the RAM config faster than AMD's Strix Halo.
https://www.amd.com/en/products/processors/laptop/ryzen-pro/ai-max-pro-400-series/amd-ryzen-ai-max-plus-pro-495.html
These really should've launched last year. I think MediaTek even shipped some phone SoCs with ARM's C1 Ultra cores, back in 2025.
Let's hope they prompt some game publishers to offer native ARM builds of a few top-tier games. Maybe that'll do some good for Windows/ARM.
P.S. let's hope we hear something about AMD's Soundwave, although the seeming lack of recent rumors isn't a good sign. -
usertests Reply
Sound Wave (with ARM cores) sounded like a Wildcat Lake or Snapdragon C competitor.bit_user said:P.S. let's hope we hear something about AMD's Soundwave, although the seeming lack of recent rumors isn't a good sign
If it doesn't materialize, I would look for Bumblebee instead. -
erazog Why would anyone buy these for games, the laptop market is overwhelming devoted to productivity.Reply
I guess the reveal and marketing push will be all about AI as thats what makes people buy Nvida Cuda for. -
ekio That’s exciting !!Reply
Finally some high end laptops not running a 50yo completely obsolete ISA that drains the battery like an oven, with an unified memory. That’s 2 serious actors, can’t wait to see what AMD will come with. -
dva852 I think the launch will be of interest to mainly AI enthusiasts.Reply
For them, N1's 128-bit memory bandwidth and 64GB cap is basically DOA.
N1X's lower-end variant, at roughly 80% compute of the top-end and same memory bandwidth, should be a sweet spot, depending on pricing. 128GB is practically mandatory.
Gaming perf would be nice-to-have, but a bit irrelevant.
The above config would be more desirable than Strix Halo as an AI box, because of CUDA. Strix Halo mini-PCs have jumped from ~$2K to roughly ~$3K (and DGX Sparks boxes from ~$3-4K to ~$5K) because of RAM shortages. If N1X 128GB config can get under $3K, it will have appeal.
More likely, however, AI peeps will wait for the mini-PCs to appear. Laptops have too much thermal constraints. -
Notton Reply
I hate to be the bearer of bad news for you with regards to N1X laptop pricing, but...dva852 said:I think the launch will be of interest to mainly AI enthusiasts.
For them, N1's 128-bit memory bandwidth and 64GB cap is basically DOA.
N1X's lower-end variant, at roughly 80% compute of the top-end and same memory bandwidth, should be a sweet spot, depending on pricing. 128GB is practically mandatory.
Gaming perf would be nice-to-have, but a bit irrelevant.
The above config would be more desirable than Strix Halo as an AI box, because of CUDA. Strix Halo mini-PCs have jumped from ~$2K to roughly ~$3K (and DGX Sparks boxes from ~$3-4K to ~$5K) because of RAM shortages. If N1X 128GB config can get under $3K, it will have appeal.
More likely, however, AI peeps will wait for the mini-PCs to appear. Laptops have too much thermal constraints.
https://videocardz.com/newz/nvidia-n1x-650-and-n1x-675-appear-in-lenovo-yoga-pro-7-listings
>€3000 for N1X/32GB, roughly US$3500
>€4000 for N1X/64GB, roughly US$4700 -
bit_user Reply
AMD's Soundwave was rumored to be relatively low-end. As @Notton said, it was to be more of a Wildcat Lake competitor. However, it might've gotten cancelled.ekio said:can’t wait to see what AMD will come with.
I'll say this: I think AMD does need to break into the ARM server market. So, it would make sense if they used a low-end SoC as a test vehicle for their own ARM cores. However, the latest rumors I heard about it were that they simply licensed cores from ARM. -
bit_user Reply
Didn't watch, but I'm just going to point out that it's not Nvidia, but rather MediaTek who seems to be the lead partner in this. They tapped Nvidia to provide the iGPU, perhaps similar to the announcement Intel made about using Nvidia graphics in unspecified future CPUs.usertests said:JdB722MK380
Nvidia does make its own ARM-based SoCs, which were originally aimed at phones and tablets, but have now gone towards embedded applications and robotics. Also, self-driving. The Nintendo Switch & Switch 2 are among the most high-profile devices using them.