IMHO the absence of in-depth Snapdragon reviews on release day does, in fact, telegraph the lack of confidence that both Qualcomm and MicroSoft have.
Qualcomm Snapdragon X (Oryon)
- Thread starter continuum
- Start date
-
- Tags
- arm cpu qualcomm snapdragon x
At least based on some of the CB2024 benchmarks for MT it looks like a M2 Pro, 96% of the performance and same system power. Seem to mainly be the lower skus currently so the top of the line ones may do a bit better.
If this had gone out early last year it would have been impressive but now it looks just decent in the windows space and a gen or two behind vs Apple. Will be interesting to see if that holds once the full reviews come out.
If this had gone out early last year it would have been impressive but now it looks just decent in the windows space and a gen or two behind vs Apple. Will be interesting to see if that holds once the full reviews come out.
Two things I've noticed after watching about three reviews this morning... it seems like a number of reviewers (at least two of the three I watched this morning) received the lowest end X Elite processor. One reviewer also called out that the power settings / minimum processor state in Windows by default is set at 40% for both battery and plugged in so he put both to 100% for benchmarking (he also is one of the ones with the lowest end X Elite). Seems most reviewers have the Asus as well.
This is the guy who talks about the power settings and has the lowest end X Elite in a Galaxy laptop:
Here's another who got a Microsoft Surface with a lower SKU (middle SKU) and a Galaxy with the highest (haven't gotten far into this one, yet):
This is the guy who talks about the power settings and has the lowest end X Elite in a Galaxy laptop:
Here's another who got a Microsoft Surface with a lower SKU (middle SKU) and a Galaxy with the highest (haven't gotten far into this one, yet):
Last edited:
I trust that Dave ran apps that aren't ARM native, but I can believe that he didn't run enough to fully test the compatibility. The long light duty battery life coupled with running Windows is compelling, but if even half of Charlie's article is correct, those benefits don't balance out all the pitfalls.
The first guy "Matt" appears to have no idea what he's talking about. Raising the minimum processor state to 100% is likely to harm, not help, CPU performance, by squandering thermal capacity on silicon that has no genuine need of being on. And there are numerous trivial inaccuracies throughout the rest of the video. Classifying "Matt" among the majority of PC youtubers who don't deserve the views.
Well... judging by the very quick glance I gave his feed, it looks like he's primary a Mac guy, so...
I've been just doing broad searches on YouTube to find stuff. Seems like most of the reviewers got the Asus thing so I haven't watched a lot of those.That Asus laptop seems like the most boring of the bunch until you realize it's $1299 with 1TB storage and a 15.6" 2k+ 120Hz OLED screen. And that's MSRP, it'll be discounted. It's not necessarily the one I'd prefer, but the value is reasonable.
Interested to see how Lenovo, Dell, Framework, MinisForum, and others implement.
Interested to see how Lenovo, Dell, Framework, MinisForum, and others implement.
Evil spin: Perhaps Asus plans to make their money with overpriced and unnecessary repairs ...
(nope, not being serious)
This is a solid review. To sum it up the
Snapdragon X deliveries solid performance but doesn't live up to the hype. The reviewer makes a good point that Intel and AMD alternatives can be purchased for less money and delivery similar performance, battery life, and cooling.
View: https://youtu.be/dT4MstOicfQ?si=Rt8RmD5Cw569z2T3
Snapdragon X deliveries solid performance but doesn't live up to the hype. The reviewer makes a good point that Intel and AMD alternatives can be purchased for less money and delivery similar performance, battery life, and cooling.
View: https://youtu.be/dT4MstOicfQ?si=Rt8RmD5Cw569z2T3
The 32G version of that is in stock here already at the local major retailer, and the price isn't too bad relative to alternatives. Interesting that it sports the no-turbo, straight 3.4ghz, X1E-78-100 chip.
Yeah, I bet that turbo mode ruins the power-sipping image for a few extra benchmark points. I'd like to see a passive machine, actually.
Yes! Major advantage to the MacBook air. You may think it's a small thing, the MacBook pro fans are super quiet 99% of the time, but in reality it's a huge selling point. Not because people are thinking about cleaning their gross fans in four years, because it's cool. No moving parts, nothing to break.
Also note the GPU is much more powerful in the higher end models, it isn't just the turbo and stock clocks.
Also note the GPU is much more powerful in the higher end models, it isn't just the turbo and stock clocks.
I made a few "Pts / GHz" charts before; thought I'd update them with the shipping #s for the X1E-80 SKU that was finally independently tested.
Note: the Arm Cortex-X4, X3, and Apple M4 scores are passive. I picked active cooling for the rest so that we can see the uArches at their best.
Sources: here and here.
Cannot remember the last time we had four high-performance CPU uArches for consumers. Ideally, Qualcomm moves into the desktop (as do NVIDIA, MediaTek, etc. for the Cortex-X925) sooner rather than later.
Some observations:
| CPU | GB6.2 1T Pts | Peak 1T Frequency | Pts / GHz | Pts / GHz % |
| Qualcomm X1E-80-100 | 2845 | 4.000 GHz | 711 Pts / GHz | 100.0% |
| Intel i7-14900K | 3243 | 6.000 GHz | 541 Pts / GHz | 76.1% |
| AMD 7950X | 2975 | 5.700 GHz | 522 Pts / GHz | 73.4% |
| Apple M3 Pro (12C) | 3138 | 4.056 GHz | 774 Pts / GHz | 108.9% |
| Apple M2 Pro | 2663 | 3.504 GHz | 760 Pts / GHz | 106.9% |
| Apple M1 Pro | 2409 | 3.220 GHz | 748 Pts / GHz | 105.2% |
| Apple M4 | 3715 | 4.400 GHz | 844 Pts / GHz | 118.7% |
| Arm Cortex-X4 (8G3 Galaxy) | 2287 | 3.390 GHz | 675 Pts / GHz | 94.9% |
| Arm Cortex-X3 (8G2 Galaxy) | 2107 | 3.360 GHz | 627 Pts / GHz | 88.2% |
| Arm Cortex-X2 (8+G1) | 1806 | 3.200 GHz | 564 Pts / GHz | 79.3% |
Note: the Arm Cortex-X4, X3, and Apple M4 scores are passive. I picked active cooling for the rest so that we can see the uArches at their best.
Sources: here and here.
Cannot remember the last time we had four high-performance CPU uArches for consumers. Ideally, Qualcomm moves into the desktop (as do NVIDIA, MediaTek, etc. for the Cortex-X925) sooner rather than later.
Some observations:
- Even with the same Chief Architect, Oryon shipped w/ lower IPC than the Apple M1.
- The Oryon core is now the highest IPC uArch on Windows.
- AMD Zen5 is projecting, at best, a +16% IPC uplift. If it applies to GB6.2, expect 3451 Pts @ 5.7 GHz and 605.5 Pts / GHz.
- Arm X925 is projecting, at beast, a ~14% IPC uplift in GB6, expect 2653 Pts @ 3.39 GHz and 783 Pts / GHz.
- AMD & Intel are reaching Arm Cortex's IPC from 3-4 years ago.
- Apple's M4 was the biggest uArch redesign since M1, even though it's just +12.8% IPC vs M1. With such a high base IPC, a relative 10% is massive (equivalent to a 20% jump at half the IPC). Apple needs +84 Pts for +10% IPC, while Intel only needs +54 pts for +10% IPC.
- As suspected, M1 -> M2 - > M3 were very minor iterations in Pts / GHz.
- No efficiency data for anything above, but everything seems to be settling at 10W to 20W / core, with Intel & AMD at the high end and Apple at the low end of that range.
Last edited:
Wow, much appreciated, @ikjadoon ! And don’t worry about not including efficiency numbers, some of the cores clearly are not aimed at the same segments, so there’s plenty of other design decisions and trade-offs made that I think including that might be… Not distracting, but perhaps an oversimplified view.
That's a really useful chart. Here's some more numbers that I looked up for my own comparison:
You can really see how the Z1E, 7840U, and 8840U are the exact same chip, at least on 1T CPU performance. The mobile CPUs show up pretty well here because they aren't particularly power-constrained using only a single thread.
Also note the geekbench browser has the X1E-80-100 as either 2538 or 2738, not 2845, but I went with your number for simplicity. It's weird, because the higher of those 2 numbers shows as a lower clockspeed and it shouldn't show as two separate CPUs either. Still early days.
| CPU | GB6 1T Score | Clock | Points/Ghz | Notes |
|---|---|---|---|---|
| X1E-80-100 | 2845 | 4.000 Ghz | 711 | |
| AMD 7800X3D | 2717 | 5.000 Ghz | 543 | best gaming desktop |
| AMD 8845HS | 2395 | 5.100 Ghz | 469 | best mobile gaming, 45w |
| AMD 5800X3D | 2094 | 4.500 Ghz | 465 | best gaming desktop, last-gen |
| AMD 5950X | 2191 | 4.900 Ghz | 447 | AMD 16-core, last-gen |
| AMD 7840U | 2243 | 5.100 Ghz | 440 | popular handheld gaming, last-gen |
| AMD 8840U | 2234 | 5.100 Ghz | 438 | best mobile gaming, 28w |
| AMD Z1-Ext | 2207 | 5.100 Ghz | 433 | popular handheld gaming |
| Steam Deck | 1153 | 3.500 Ghz | 329 |
You can really see how the Z1E, 7840U, and 8840U are the exact same chip, at least on 1T CPU performance. The mobile CPUs show up pretty well here because they aren't particularly power-constrained using only a single thread.
Also note the geekbench browser has the X1E-80-100 as either 2538 or 2738, not 2845, but I went with your number for simplicity. It's weird, because the higher of those 2 numbers shows as a lower clockspeed and it shouldn't show as two separate CPUs either. Still early days.
Last edited:
This seems very odd. Restricted access to some critical IP, or different priorities in the design?
Even with a similar architecture, it's a different implementation. I haven't tried to tally the differences but I'm betting the M1 and the Oryon are significantly different implementations of even similar designs so there will be differences in operation/performance. But yeah... probably some differences in priorities and such as well, which will change designs and implementations.
Geekbench has a history of measuring the OS along with the CPU+compiler combination. For any given 'x86 machine, scores under Windows are usually notably lower than under Linux. Presumably the unix based MacOS is also looking a bit better here than Windows on ARM.
Ordered a Yoga Slim 7X with 32GB of RAM last night.
My goal is to never boot Windows on it, though. Supposedly Qualcomm has been working on upstreaming a bunch of the drivers already, but I guess I'm about to find out firsthand.
Ooh man, I would definitely wait for reviews on the new AMD chips coming out extremely soon, if not Intel with arrow lake.
Status updates are kept here:
Upstreaming Linux kernel support for the Snapdragon X Elite
Latest SoC is getting traction on Windows-based laptops, and we’re also consistently upstreaming to mainline Linux. See areas of focus and try our Debian installer.
IIRC this was due to the choice of memory allocator on each platform rather than anything OS related.
Probably due to the much higher clock speed. It is hard to have high IPC when targeting high clock speed since you have less time per cycle (meaning instructions take more cycles to complete) and your memory access times (measured in clock cycles) become longer. Taken to the extreme you have things like DSP cores where you run at a few hundred MHz, cache is zero latency and DRAM accesses are almost free and so you can have extremely high IPC (but lower overall performance since clock rates are so low).
If you can offer any authoritative source I'll gladly accept it.
I can't provide one, there were some threads on RWT about this a few years back where this discrepancy came up, and IIRC the subtests showing the largest differences were the tests that would be more sensitive to memory allocator performance.
At the very least that hypothesis makes more sense that MacOS or Linux having some sort of inherent advantage over Windows on a test that isn't meant to exercise system performance.
I would not be surprised. Microsoft at least has some experience with the past Surface Arm devices and thin, high performance (relatively) devices, etc. Dell is probably treating it as a 'let's see what happens and if people want it' kind of situation rather than going all in to produce a really good option. If the Surface one is good enough to attract buyers, Dell will eventually ape their designs a bit and produce something that will end up in corporate fleets by default. I don't think they care about marketing it.
The surface laptop is a beautiful device, and reasonably priced. I'd still definitely wait to see how AMD strix point (I refuse to use that stupid AI name) performs. Rumors are it's extraordinary.
According to https://blog-hjc-im.translate.goog/...l=auto&_x_tr_tl=en&_x_tr_hl=fr&_x_tr_pto=wapp Strix Point is a 10% IPC improvement over its predecessor in SpecCPU (and a 22% improvement in ST).
Also interesting to note that AMD has adopted the clustered decoding approach of Tremont and its successors, which exploits the fact that a branch target always marks the start of a new instruction.
I’m personally more interested in efficiency than performance in mobile, particularly the GPU which is rumored to be much faster. Want to see what zen5c can do.
It will be interesting to see what happens when AMD and Intel's new offerings are released. Even if they only offer a similar level of performance, they'll still have an advantage because they don't rely on an emulation layer for x86 apps. With Apple's M processors, you knew that you were future-proof because they're doing away with x86 support. Microsoft doesn't have that luxury since they can't or won't break legacy support.
Yes my impressions have somewhat solidified as reviews release. Barring any "magic drivers" (which are totally possible!) these laptops are only really interesting if you strongly prioritize battery life. And who knows, maybe AMD or even Intel (heh) will surprise us. Intel's working on new architectures also, in both CPU and GPU actually, and they likely have tons of low hanging fruit to pluck too.
My feeling is they were priced too high. Their pricing looks great compared to AMD until you remember that some stuff will be wonky, and probably it'll take a year or two to cover the vast majority of applications, while the long tail, legacy hardware support, many older games, etc, will never be really covered. And all that time the competition isn't standing still.
These are going to be killer bargains when they get steeply discounted. I'd pay $600 for that Asus laptop any day of the week and count myself happy. Put debian arm and KDE on it. Wouldn't be surprised if it gets there in maybe eight months.
My feeling is they were priced too high. Their pricing looks great compared to AMD until you remember that some stuff will be wonky, and probably it'll take a year or two to cover the vast majority of applications, while the long tail, legacy hardware support, many older games, etc, will never be really covered. And all that time the competition isn't standing still.
These are going to be killer bargains when they get steeply discounted. I'd pay $600 for that Asus laptop any day of the week and count myself happy. Put debian arm and KDE on it. Wouldn't be surprised if it gets there in maybe eight months.
Last edited:
Are you even able to install other OS's on it? Windows on ARM laptops have often been boot-locked.
Don't Intel laptops already have much better battery life than AMD? Or did things switch around with the latest chip generation?
[... about permission to boot Linux on Windows/ARM devices ...]
As far as I know, Microsoft can revoke the required certificates any time. That won't brick existing devices immediately. But any connection to the internet could pull in a firmware upgrade for the respective TPM.
I don't think it is all too likely that MS would ever act so bluntly. But Linux on ARM seems to depend somewhat on MicroSoft's mercy. Even more so now that MicroSoft's own TPM is mandated ("Pluton").
IIRC, AMD performs much better under load but still don't idle as well as Intel. I think AMD's been ahead while doing things since Zen2 vs. Tiger Lake.