Nvidia N1X leak points to limited 2026 availability

[Redaktion]

Moore's Law is Dead has revealed Nvidia plans to show off its N1X laptop CPU at Computex this year. However, one might have to wait a while before they can buy a laptop powered by the chip.

https://www.notebookcheck.net/Nvidia-N1X-leak-points-to-limited-2026-availability.1282855.0.html

[opckieran]

There's NO WAY this performs close to a desktop 5060 with essentially half the power budget.

[._.]

There's NO WAY this performs close to a desktop 5060 with essentially half the power budget.

You do realize this has 6144 CUDA Cores compared to a desktop 5060's 3840, right?

Not that any of this matters as unfortunately it's looking like this is a flop. :/

[Huang]

Nvidia is clearly not targeting the hardcore standalone RTX5070 GPU desktop gamers, Jensen is going for the laptop users, and probably gamers AI developers who are also in the Mac ecosystem.

[yee]

There's NO WAY this performs close to a desktop 5060 with essentially half the power budget.

The chip will be support up to 128 GB of LPDDR5x-8533 RAM and be manufactured on an unspecified 3 nm node from TSMC.

Gaming performance is based on memory bandwidth, let's calculate it:
N1X: 273 GB/s = 256-bit * 8533 MT/s / 8 / 1000.

Strix Halo: 256 GB/s = 256-bit * 8000 MT/s / 8 / 1000.

Basically the N1X falls into the 256-bit APU (e.g. Strix Halo (=4060 Mobile/Laptop performance)) category.

the same as the DGX Spark

Another confirmation. en.wikipedia.org/wiki/Blackwell_(microarchitecture): same GB10 chip.

3dmark.com/search - Steel Nomad:
Strix Halo' Radeon 8060S iGPU: Average score: 2031
RTX 4060 (notebook): Average score: 2262

Memory bandwidth of a 4060 Laptop is 224 - 256 GB/s (aligns perfectly with Strix Halo' gaming performance).
Memory bandwidth of a 5060 (Ti) desktop is 448 GB/s (64% more than N1X).
Memory bandwidth of a 5050/5060/5070 Laptop is 384 GB/s.
(en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units)

Based on the memory bandwidth alone, I'd agree with you. But maybe due to it being based on TSMC' 3nm node variant, a bit more of performance can be squeezed out.

But between "between an RTX 5070 and RTX 5070 Ti laptop", would be very surprising, if N1X' memory bandwidth is at 273 GB/s.

[noo]

Gaming performance is based on memory bandwidth

Only at higher resolution gaming at 4k and beyond. You don't need much for lower resolutions (e.g. 720p-1080p)

Blackwell is slightly more efficient at data compression than RDNA3.5 too.

Basically the N1X falls into the 256-bit APU (e.g. Strix Halo (=4060 Mobile/Laptop performance)) category.

Yeh, because games aren't compiled for arm and there is fare bit CPU overhead.

If a game was ever released for arm, it's going to be faster.

Wendell from level1tech has tested all these systems (DGX Spark GB10, Halo) and said pretty much the same.

Maybe laptop implementation will be slightly slower because less power. DGX Spark can use over 100+ watts.