Aardvark Daily

Please visit the sponsor!

Smoke, mirrors and a leather jacket

Earlier this week I reported on NVIDIA's big announcement at Computex -- the "new" unified CPU/GPU it calls the N1X that is a "smartphone moment", according to the company's head.

By integrating up to 20 ARM cores and a CUDA-based GPU that is roughly equivalent to an RTX5070, the N1X silicon is brand new and will change the face of portable and desktop computing -- so we were told.

Well it seems that Mrs Huang's boy has been not completely honest with us. In fact, some say he's been trying to pull the leather (of his trademark jacket) over our eyes.

Is the N1X really brand new? Is it going to revolutionise the market?

Let's take a closer look and find out.

If the N1X is brand new, why does it use older ARM cores than even Apple's latest M-series chips?

The official line was that these older cores would be more energy-efficient.

The reality is that they're probably not. They're just older and that's because the N1X is not as new as Ms Huang's boy would have us believe. Little Jensen, it would appear, has been somewhat conservative with the truth.

People much smarter and more observant than me have noticed that the "all new and very shiny N1X" chip is nothing more than (perhaps a variant of) the same silicon that has been used in the DGX Spark product first announced back in March 2025 and which has been shipping since the 4th quarter of that year.

I've seen videos which show that the piece of silicon being proudly touted by Jensen at Computex 2026 has exactly the same die number as that found in the GB10 silicon used in the DGX Spark systems that have already been shipping for over nine months.

In effect, it appears that NVIDIA have taken their enterprise product (the GB10) and relaunched it as a consumer-grade one (the NX1) -- amidst much fanfare and hype.

There may be some tweaks involved and some commentators are speculating that one of the biggest benefits to NVIDIA of this new direction is that it will allow them to repurpose all the flawed silicon that would otherwise be rejected for GB10 use. By way of a binning process, GB10 silicon that has faulty CPU or CUDA cores can be resold as lesser versions carrying the N1 label. We've been told that the N1 variants will have fewer cores so that makes sense.

Now NVIDIA can turn their scrap into cash.

This also plays very nicely into NVIDIA's long-term vision of the computing marketplace. They want people to run "thin clients" rather than more capable systems that can be upgraded and expanded as computing needs grow. The unified RTX Spark concept (Windows machines powered by the NX1 or N1 chips) fits that description perfectly.

If Mrs Huang's boy has his way, we'll all be running "Windows AI" on his N-series chips which connect to NVIDIA-powered cloud servers when the real heavy-lifting tasks are required.

Welcome to the future and that future is green -- at least according to jacketman.

Carpe Diem folks!

Please visit the sponsor!

Here is a PERMANENT link to this column

Rank This Aardvark Page

Change Font