Discussion Nvidia N1X ES Geekbench score surfaces

Jul 27, 2020
28,173
19,203
146

1741096765720.png

The score is exactly what I would expect from a Mediatek chip.
 
  • Like
Reactions: lightmanek

gdansk

Diamond Member
Feb 8, 2011
4,689
7,927
136
That's pretty low. Even if all Cortex-A series cores. Are there any rumors about the configuration?
 

LightningZ71

Platinum Member
Mar 10, 2017
2,648
3,327
136
That single core is Dimensity 8100/8300 range (A78-A715 around 3Ghz).
That multicore is Dimensity 7050-7200 pro range, (2 X A78 or 2 X A715 + 6 X A55/A510)

With the reported metadata showing a quad of A725, there's obviously artificial limits on whatever system this was as that core at those speeds should be capable of more...
 

jdubs03

Golden Member
Oct 1, 2013
1,350
943
136
That single core is Dimensity 8100/8300 range (A78-A715 around 3Ghz).
That multicore is Dimensity 7050-7200 pro range, (2 X A78 or 2 X A715 + 6 X A55/A510)

With the reported metadata showing a quad of A725, there's obviously artificial limits on whatever system this was as that core at those speeds should be capable of more...
So are we saying that they’re not using any X925s?

If not, what target market are we even looking at here. They need at least Dimensity 9400+ performance to be worthwhile.
 

LightningZ71

Platinum Member
Mar 10, 2017
2,648
3,327
136
I don't know what, if any, separate prime core(s) they will have. All we know is that someone caught the metadata and it reported 4x725. With the test reporting two clusters of 2 processors, at least 2 of them are lower performance. I still think this is a test platform of some sort that isn't operating at full spec.

For performance, note that Nintendo expects to be competitive with 4xA78C cores for the Switch 2(wikipedia). The Shield still does surprisingly well with 4 x A55.

For longevity and absolute performance reasons, I hope that they have at least 2 cores at Max spec. A pair of A725 with max frequency and cache config with at least 6 mid-performance cores to handle support threads would be just fine for most anything you're going to throw at it.
 

Nothingness

Diamond Member
Jul 3, 2013
3,354
2,441
136
That's not what I expected.

I expected something much faster than the 9400, which is much faster than what you posted.

View attachment 119057
The picture posted shows an ID of 0xD87 which is 3463. This is indeed Cortex-A725: https://developer.arm.com/documenta...-registers-summary/MIDR-EL1--Main-ID-Register

So this should be compared to something like this: https://browser.geekbench.com/v6/cpu/10880882
The result is for a 3 GHz CPU, so I guess the NVIDIA result is for a 2 GHz CPU.
 
Jul 27, 2020
28,173
19,203
146
I find this SOOOOO funny!


Nvidias N1X will steal the show while being sheeet​

One of the major concerns the whole industry faced was Nvidia's known aspirations to enter the mobile market. Now we can all take a deep breath (yes, also you, AMD and Qualcomm investors out there) that this product is by no means anything special. Yes, for sure, this product will get the most headlines and will be totally overhyped, simply because it's Nvidia and Jensen. I was extremely scared once even Michelle Holthaus confirmed a new entry into the market (Nvidia) that we could be dealing with a product that will shatter everything Intel can offer. There I witnessed my own fall for the absolute marketing and Wall Street dominance Nvidia has. We just presume everything they release is made out of diamonds and better than everything else. This CPU shows... let's say... that they have become a bit fat and lazy. Let's explain why this product is not important to us:

N1X is a mobile CPU with 10P+10E ARM Cortex Cores that will have the RTX 5070 as an iGPU in it. It is in fact a variant of the GB10 Superchip and it will release in 2026. There are multiple rumors circulating for months that Nvidia has some sort of technical issues with N1X, and the launch is getting delayed repeatedly. One very important factor everyone needs to take into account is that the N1X will be a premium chip that will be priced very high. This is not a product for the mass market, but let's see how our best Panther Lake could perform against it.


SKU Single Geekbench Multi Geekbench Single CinebenchR24 Multi CinebenchR24 OpenCL Score for Xe Graphics
N1X 10+10+6144CUDA 3'090 18'800 unkown unkown 46'300
PTL-H 4+8+4+12Xe 3'080 17'440 143 995 46'950
PL2 (don't confuse it with PL1 please; this is not the base power) of Panther Lake is 64W; this thing will have 125W. So... I think the point comes across. It will be an inefficient, highly expensive piece of Jensen's greed. Sure, it will be slightly more performant, but it will also be priced much higher while consuming 60-100% more power.