News Early Snapdragon X Elite benchmark shows Arm CPU is faster than AMD's top-end mobile APU

Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Is this any surprise given what we saw with the Apple M1? Actually this chip gives the same overall multi-core result as the M1 in Geekbench, and is only about 20% slower than the M3.

The real question, however, will be how well it runs x86 programs in relation to native x86 processors. If the translation layer is as seamless as Apple's then there's no problem, but if it's not, then there's going to be a major issue, especially since Arm based Windows laptops aren't exactly cheap compared to x86.
 
It looks good for now, but I wonder how it will scale.
For example, the M1 was stellar when it launched, but M2 and M3 were lackluster improvements comparatively.

Computation benchmark numbers are one thing, but I'd like to see the power consumption/efficiency numbers as well.
 
  • Like
Reactions: bit_user
IMO, these CPUs just need to be fast enough. Obviously, faster is better, but as long as they're not significantly faster or slower, minute differences in performance won't sway people one way or another.

I predict Qualcomm will shoot itself in the foot by pricing these like they're Apple products, however. I think they have consistently undermined their previous attempts to penetrate the Windows laptop market through unrealistic pricing, and now that they have something actually competitive, I expect them to double-down on this unsuccessful strategy.
 
  • Like
Reactions: thestryker
Alvar "Miles" Udell said:
The real question, however, will be how well it runs x86 programs in relation to native x86 processors. If the translation layer is as seamless as Apple's then there's no problem,
Click to expand...
You can look at reviews of the Lenovo Thinkpad X13S. That's a Win11/ARM laptop, based on the Snapdragon 8cx Gen 3. I think I recall NotebookCheck's review saying they had compatibility problems with a few of the programs they tried to run, but it was basically the first laptop to support Win11/ARM and that's when Microsoft launched their emulator for x86-64. Prior to that, they only supported running 32-bit x86 code on ARM. So, one might hope those early issues have been sorted out, to some extent.
 
bit_user said:
You can look at reviews of the Lenovo Thinkpad X13S. That's a Win11/ARM laptop, based on the Snapdragon 8cx Gen 3. I think I recall NotebookCheck's review saying they had compatibility problems with a few of the programs they tried to run, but it was basically the first laptop to support Win11/ARM and that's when Microsoft launched their emulator for x86-64. Prior to that, they only supported running 32-bit x86 code on ARM. So, one might hope those early issues have been sorted out, to some extent.
Click to expand...
And according to a How To Geek article from last week the translation layer currently available is still an effective steaming pile of poo compared to Apple's Rosetta 2 translation layer. But since this processor won't be out until likely 24H2 time, a supposedly massive overhaul of Windows 11 will hopefully improve the translation layer if Microsoft has any hope of making Windows On Arm, and potentially a new fully featured Windows Phone, a reality, or a joke.
 
  • Like
Reactions: kealii123
To be honest, the results are not surprising given that the M1 already demonstrated that ARM based chips can be very potent, especially under tight power constraints. But I do think the performance is not that good considering this chip contains 12 performance cores as compared to say M3 or M3 Pro. The other thing to consider is software optimisation. In benchmarks, that can be easily optimised and may represent the best case scenario.
 
watzupken said:
I do think the performance is not that good considering this chip contains 12 performance cores as compared to say M3 or M3 Pro.
Click to expand...
On a laptop power budget, you might get better multithreaded performance to have some E-cores in the mix. They provide better perf/W, which is key when you're power or thermally -limited.

watzupken said:
The other thing to consider is software optimisation. In benchmarks, that can be easily optimised and may represent the best case scenario.
Click to expand...
Yeah, without looking at how predictive GeekBench is of performance on other workloads, which can also vary between x86 and ARM, it doesn't necessarily tell us much.

That's also a big downside of benchmarks that are closed-source. With open source benchmarking suites, like Phoronix Test Suite, you could look at the source code of a given benchmark and see if it has hand-coded assembly or intrinsics for one ISA vs. another. SPECbench is a little similar, in that it's delivered as source code and includes some open source programs.

Speaking of different benchmark scores, I ran across this performance preview article that Anandtech published last year (too bad it doesn't contain values for any x86 or Apple laptops, for easy comparison!).

That was already 4 months ago! If they had silicon mature enough for performance demonstrations, I can't help but wonder why these aren't already shipping!
 
  • Like
Reactions: usertests
Just to be clear, geekbench is more indicative of cache / memory performance than CPU, and also it can't be used to compare efficiency simply because CPUs don't hit their power budgets, its' a relatively light workload.
 
  • Like
Reactions: jlake3
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom