Nuvia: don’t hold your breath

Looks like we have the first X Elite Geekbench score.

Edit: Actually there’s quite a few scores there now and all the scores over 3000 are on Linux where we know the fans were on full blast.
That's a good score. It will still be very competitive next year
 
Who knows what next year will bring? At best we can say it’s fine for this year, depending on the power used to achieve it.
It's also fine for the PC market next year. Intel won't be close this anytime soon. More like 2025 with Lunar Lake assuming it's actually powerful like X Elite.
 
Geekbench Linux aarch64 (not a supported release, it’s in preview) always reports 1 core/n threads.
Raspberry Pi 4/5, Honeycomb LX2, Graviton instances, Linux VMs on Mac (Parallels/VMWare) etc. doesn’t matter, always reports 1 core.
 
Those recent Geekbench results appear to be the lower clocked 23W TDP SKU (assuming the reported clock speed is correct).

23W = 4GHz 2-core turbo, 3.4GHz all core turbo
80W = 4.3GHz 2-core turbo, 3.8GHz all core turbo

Without knowing what TDP means in this case, we can’t really judge it against the competition.

Core Ultra 9 185H has a maximum turbo power of 115W(!!). The bursty nature of Geekbench means it probably gets through a run before getting throttled back to 65W.
 
Another thing.

There’s something off about Geekbench for Windows aarch64. I don’t know if it’s a Windows issue or a Geekbench issue (different compiler tuning?), but it scores lower than macOS and Linux

Here are some M3 Pro results I took last year
macOS native = 3176 points https://browser.geekbench.com/v6/cpu/3472954
Linux aarch64 (VMWare) = 3039 points https://browser.geekbench.com/v6/cpu/3480434
Windows aarch64 (VMWare) = 2755 points https://browser.geekbench.com/v6/cpu/3479543

IIRC, the early X Elite Windows and Linux results had a similar performance delta
 
Another thing.

There’s something off about Geekbench for Windows aarch64. I don’t know if it’s a Windows issue or a Geekbench issue (different compiler tuning?), but it scores lower than macOS and Linux

Here are some M3 Pro results I took last year
macOS native = 3176 points https://browser.geekbench.com/v6/cpu/3472954
Linux aarch64 (VMWare) = 3039 points https://browser.geekbench.com/v6/cpu/3480434
Windows aarch64 (VMWare) = 2755 points https://browser.geekbench.com/v6/cpu/3479543

IIRC, the early X Elite Windows and Linux results had a similar performance delta
Huh I thought they fixed that.
 
Tech sites really go out of their way to give Qualcomm a pass. 1628 is not a respectable number and I doubt it’s even a real number. 2574 is fine, but again, nowhere near where they pretended to be and it is not “pretty close” to 3181. The M3 is 24% faster. Bad journalism.
techradar-qualcomm.png
 

Attachments

  • techradar-qualcomm.png
    techradar-qualcomm.png
    443.7 KB · Views: 15
Tech sites really go out of their way to give Qualcomm a pass. 1628 is not a respectable number and I doubt it’s even a real number. 2574 is fine, but again, nowhere near where they pretended to be and it is not “pretty close” to 3181. The M3 is 24% faster. Bad journalism.
View attachment 28591


“Which narrowly beats it.”
 
In message #20 on this thread I said “I suspect we will see M1-like performance, for what it’s worth, but in 2023 that may not be good enough.”

Looks like it’s more like M2-like performance in 2024. It’s actually not bad, though I want to see power numbers and more details. But, when the thread was started, we have to remember how these guys were being billed as “apple-killers.”

They may very well be able to beat Apple at some point, but I think it will eventually settle into a leapfrogging situation.
 
Back
Top