It'll be interesting to update this chart when M3 chips come out so we can compare them.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Qualcomm revealed X Elite's benchmark scores
- Thread starter sunny5
- Start date
- Sort by reaction score
First Geekbench results for the M3 Max are out and it hit 21084. That’s 37% faster than the Elite at 80W.
Mac15,9 - Geekbench
Benchmark results for a Mac15,9 with an Apple M3 Max processor.
browser.geekbench.com
That's really good, now what is it compared to just the base M3?
I decided to check:
Snapdragon X Elite 23W: Single: 2780 Multi Core: 14029
Snapdragon X Elite 80W: Single: 2971 Multi Core: 15372
(which means basically buy devices with the 23W honestly...)
Now, the M3 line:
base M3: Multi Core: 11700 - Oh, only 83% as fast compared to the 23W.
M3 Pro: Couldn't find it yet
M3 Max: Multi Core: 21084 - NICE!
So, what we discover is that there is give-and-take here for both sides.
Windows will be able to more than hold its own with the base M3, and probably the M3 Pro. This is really big and should not be overlooked.
And Apple gets to show what years of experience in the ARM world can do by achieving 37% increase over the Snapdragon at 80W which is also huge for the mac market.
Of course, again, (this poor dead horse) Windows doesn't have to compete with the M3 Max with the Snapdragons. There are already monster machines in this market segment.
The next year is going to get very interesting. Are there diminishing returns down the ARM path into the future? Are chiplets going to make Apple have to rethink things? Either way, Snapdragon is impressive--and so is Apple's M-series.
Last edited:
Again, SD did it again as in the past years. 7501 for gen3 in multicore in open envelope, and within an smartphone did 6750, and this goes for windows on arm for 6-7 years. Those numbers will not translate when we will get the devices from dell samsung hp etc
In the meantime we know those M3 numbers came from an actual device and this we will get from the actual device as a customer
So it’s fair to say that we can shave around 10% to come to a more realistic performance
In the meantime we know those M3 numbers came from an actual device and this we will get from the actual device as a customer
So it’s fair to say that we can shave around 10% to come to a more realistic performance
Unless you have some proof of it, it is FUD. Are you suggesting that Anandtech's figures are false?
No, but historically actual devices have scored lower than Qualcomm’s sample devices. The Linux scores were high due to full blast fan use according to Anandtech.Unless you have some proof of it, it is FUD. Are you suggesting that Anandtech's figures are false?
I’m sure the hardware performance is nice, but I’m very skeptical when it comes to software. Sure, Microsoft have Office complied for Win ARM, but Beyond that? The whole Windows market has always been extremely reluctant to cut compatibility or really anything from the past.
Also, Intel seem to work very hard a getting decent CPUs out for laptops that doesn’t eat power like crazy. I’m just not convinced that even a very good ARM CPU like this will finally make Windows on ARM a success.
Also, Intel seem to work very hard a getting decent CPUs out for laptops that doesn’t eat power like crazy. I’m just not convinced that even a very good ARM CPU like this will finally make Windows on ARM a success.
Agreed, Windows 12 arm will probably be a key part of the equation, and app compatibility will have to improve, but I used a Windows 11 ARM device and was extremely surprised at how much worked already.
I disagree with this idea of an arbitrary 10 percent, but again the nice thing is that we will eventually have devices with this architecture to see how good or bad it is.
They already disabled ARM32 going forward. And I was very surprised at the level of good ARM64 apps that already worked. And when they didn't it really wasn't that much of a performance hit when I had to use x64 apps. I was pleasantly surprised (this was a Robo and Kala Surface type device). So again, I agree the proof will be in the proverbial pudding, but this is very exciting!
It’s not always easy to find evidence after the fact as these references devices don’t always have many traceable public tests.
Nonetheless here is an article where they reference how the samples usually perform differently to the final product
https://www.androidauthority.com/snapdragon
They refer to the fact that the 8gen1 products performed more slowly than the sample.
Here is another article listing some scores of the Reference 8gen2.
https://uk.pcmag.com/mobile-phones/143952/qualcomm-snapdragon-8-gen-2-benchmarks-show-promise
They list Geekbench 5 and Antutu. Their Geekbench score is 1489 single core. If you look for scores you can see most are slower than these ones. I haven’t had a chance to look for the antutu ones. From my recollection it was a little slower.
Take it or leave it.
We’ll know in a year for sure. We already have proof that their headline single core score requires the fans at full blast under Linux.
I'm not surprised if they couldn't get the same score as Qualcomm.
Qualcomm Snapdragon 8 Gen 2 Benchmarks Show Promise
Benchmark tests conducted on a reference design powered by the Qualcomm Snapdragon 8 Gen 2 show gains over last year's chip and even the Apple A16 Bionic processor.uk.pcmag.com
Yet doesn't under windows. Interesting. Also, the Snapdragon Robo and Kala I had didn't even have fans, so?
Does Windows Subsystem for Linux (WSL) work on an Arm windows machine and if so, does it work well?
Haven't used Windows much in ages, but if apple-silicon like devices start flowing with reasonable memory/ssd upgrade prices, I can see picking one up...
Honestly when I had the device, I didn't try WSL, I did try Windows Subsystem for Android--and that worked fine. Well, except that you had to use the Android App Store from Amazon. Ugg. But it performed fine.
I really only used Kindle for Android, I didn't really try anything else though due to the App Store choice.
Doesn’t what? You missed my point entirely. They boasted about their 3220 geekbench score. That score was on Linux with fans at full blast. There isn’t a single score on windows over 3000. My point isn’t that Windows needs fans, it’s that their score is misleading.
Okay, great point. So instead of 3220, we get to go with what was provided on Windows. I think we can make that work.
We have NO IDEA what QC will charge for Snapdragon Elite.
Area-wise it *appears* to be about 180mm^2, so about 60% the area of an M2 Pro (and probably about the size of an M3 Pro). On the other hand, QC is not magic – that smaller size must mean something(s) are missing; perhaps in the GPU area.
It may be a great chip for people who want to compile a lot while on wall power, but rather less desirable for gamers?
Not to mention the desirability of ARM Windows remains unclear. My personal suspicion is that, like Apple, most people who use Windows mainly use MS stuff and one or two other apps, so there's no real love or need for x86. But plenty of people claim otherwise.
Both the MS side of things (Windows RT FFS, or the ongoing slowness in 64b support) and the lame hardware side of things have allowed both sides to agree that WinArm has not taken off because there's been no reason to buy a product. Now there will be a reason to buy a product (for SOME users, maybe not for those who care about GPU?) – will we see a pent-up demand or a fizzle?
My suspicion (MS being MS, QC being QC, and PC vendors being PC vendors) the first wave of devices will come with a number of stupid limitations, bugs, and oversights (just like Apple, but without the expectation that Apple will fix most of this over the next three months). Probably not enough to end the ARM/QC/Windows experiment, but maybe enough to dampen sales below what were expected.
I expect rev 2 to be much better products – but right now we're discussing the rev 1 product against M3.
yes, because that article is not written by an objective editor/editors
Dont get your high hopes on this SD x elite....on paper is around M1 Pro (on gpu level) and around M2 on cpu in the end product
M2 Macbook Air will be far ahead of an actual fan-less 13" Laptop with SD x elite. M3 family is not in question. The SoC seems allright for an low mid tier laptops but because windows emulation/translation layer and developers that are not quite ready even after 6 years....these will not perform as well as you all expect
Bottom line, dont buy this until you tested yourself, or buy from a market that has good return policy
Last edited:
I had all of that with my first Surface Pro X (charts and numbers were looking fine at the event but..)....not anymore...until developers and MS are full in on this, sorry but this kind of hybrid of OS, SoC and software app developers will always be far too faulty IMO especially for pro users