News Nvidia N1X SoC leaks with the same number of CUDA cores as an RTX 5070 — N1X specs align with the GB10 Superchip

[Admin]

Nvidia’s N1X SoC, featuring a 20-core CPU and 6,144 CUDA cores, has surfaced on Geekbench, matching the specifications of the GB10 Superchip. While its early performance trails that of the RTX 5070, the chip’s ARM+Blackwell design and 120W TDP hint at a future where integrated graphics can rival mid-tier GPUs.

Nvidia N1X SoC leaks with the same number of CUDA cores as an RTX 5070 — N1X specs align with the GB10 Superchip : Read more

 

[Flayed]

You have the sentence that starts with "Even in this state" repeated.

 

[Amdlova]

AMD and Intel is ded

 

[jlake3]

Software support and pricing is going to make or break this thing. The iGPU on Apple's "Max" chips has theoretically been amazing for a mobile platform ever since the M1 Max, but the catalog of Apple Silicon optimized AAA titles has been too narrow and the price of hardware too high for the M4 Max MacBook Pro to get any gaming recommendations.

If laptops with these chips come in at $2,000+, but you can only play Nvidia-sponsored new releases natively and older games end up stuck with a translation layer that isn't absolutely top-notch... that's not gonna be great. And if these are only a small sliver of the Windows gaming market, getting people to optimize and QA for Windows-on-ARM without an Nvidia sponsorship may be hard.

 

[usertests]

AMD and Intel is ded

In the Geekbench OpenCL benchmark, Strix Halo 8060S is "between" the 5070 and the early N1X result. Strix Halo is 94% faster than the N1X, and RTX 5070 is 106% faster than Strix Halo. The N1X is going to get better than that, but how much?

https://www.tomshardware.com/tech-i...ip-revisions-and-weakening-demand-also-blamed

We have a bunch of blame going around for why the N1X is not coming soon. Eventually, it will be competing with faster Medusa Halo (2027?).

 

[Mr Majestyk]

So Moore's Law is Dead was right. 6 months ago he said Nvidia insiders told him they were going for AMD's jugular and wanted to dominate Strix Halo. If these specs are in final silicon, this would really put a cat amongst the pigeons, even more so with strong rumours Medusa Halo is cancelled. Qualcomm should be even more worried as their iGPU even in second gen Oryon will be light years behind. As much as I loathe Nvidia, this sounds exciting. This would also be a monster for productivity. I hope by the time of release, Windows for ARM is more polished and the vast majority of apps I would use run on ARM natively, rather than rely on emulation.

 

[Pierce2623]

AMD and Intel is ded

I really don’t see this chip hugely affecting AMD at all. It will require emulation/translation layers to run anything that’s x86. ARM won’t have a chance of taking the client PC market until they make an effort get the big publishers to get most modern AAA games running natively on ARM. Emulating PC games on newer high powered ARM devices is gaining a lot of popularity, but it’s silly to burn Steam Deck levels of power to get half the Steam Deck’s performance on better than Steam Deck hardware because you’re playing through an x86 to ARM emulation layer with DXVK bootstrapped to the top because nobody gives phone GPUs DirectX compatibility, even the damn literal RDNA GPU Samsung buys in from AMD doesn’t have DieectX support.

 

[aberkae]

Software support and pricing is going to make or break this thing. The iGPU on Apple's "Max" chips has theoretically been amazing for a mobile platform ever since the M1 Max, but the catalog of Apple Silicon optimized AAA titles has been too narrow and the price of hardware too high for the M4 Max MacBook Pro to get any gaming recommendations.

If laptops with these chips come in at $2,000+, but you can only play Nvidia-sponsored new releases natively and older games end up stuck with a translation layer that isn't absolutely top-notch... that's not gonna be great. And if these are only a small sliver of the Windows gaming market, getting people to optimize and QA for Windows-on-ARM without an Nvidia sponsorship may be hard.

For me if Nvidia decided to develop thier own OS or use Steam OS as primary or secondary OS would actually be more exciting than having to rely on Windows for growth for maximum low level efficiency. Will put Microsoft on notice and might start to threaten Apple as well.
Crazy how Intel is ironically about to really be in the rear view mirror if it doesn't get it's footing right. Also this might cause AMD to upgrade it's APUs to a higher rdna/udna nomenclature than the 3.5 version in current lineup which is also inferior to the ai capabilities of Intel's mobile apus xess and frame gen found in the ultra 258v. By the time this comes out Intel will surely have a successor to Lunar lake 288v or whatever the direct competitor chip is.
While the gpu side obviously Nvidia has the lead on all but when it comes to the cpu its a tough competition. Nvidia needs to get this right and start swinging out the gate. Hopefully the delay gives all players time to fortify their positions accordingly for some healthy competition in this space.

 

[Amdlova]

The big "AAA" almost stuck with the gamergate 2.0
Intel is almost dead...
Only AMD making chips every one will jump to ARM cpus. Server/enterprise level has 10% market share, Amd has 27% ish.

https://www.tomshardware.com/pc-com...-50-percent-of-data-center-cpu-market-in-2025

Nvidia will make everything to take a big piece of this pie.

 

[purposelycryptic]

Sounds great... For a new Nvidia Shield TV release. I'd definitely buy that, even if the price was quite ugly.

As a PC, though?

It's still Windows-on-ARM, a platform desperately trying to justify its existence in a scenario where the only benefit it offers users is a minor advantage in power efficiency over x86, at the cost of significant and unpredictable compatibility issues with its x86 emulation, and zero incentive for developers to go out of their way to waste their time and money making a native port.

No one asked for this - it continues to be the product of corporations joining hands to create something new to make money off of, rather than coming up with a a solution for an existing hole in the market. It's a solution in search of a problem, and always will be.

The technology itself is cool, it's just such a shame it is getting wasted on something so pointless.

 

[Notton]

I don't think N1X will dethrone Strix Halo.

From a performance viewpoint, both N1X and Strix Halo have memory bandwidth constraints when compared to a 5070 or 9070XT.
5070 has 672GB/s, while GB10 (N1X) has 273GB/s. That's less than half.

From a software perspective, Strix Halo works better on Linux than Windows, but they both work.
Where as N1X is struggling with Windows on ARM, but it might support Linux out of the box, since DGX spark is supposed to ship with Linux. And then there is the Mediatek equation. Their drivers leave a lot to be desired in both Windows and ARM. They are particularly bad when it comes with Wifi and BT signal quality.

From a pricing perspective, DGX Spark costs $4000 while GX10 Ascent is $3000
Strix Halo 395 desktop from Framework is $2000 baseline.
At >$3000, and especially >$4000, you might want to look into a System76 Ampere based system.