A First Take a look at What’s to Come

Final week we noticed the announcement of Qualcomm’s first post-Nuvia laptop computer SoC design, the Snapdragon X Elite. A brand new model title being anchored by a brand new, customized Qualcomm CPU core, the Snapdragon X Elite shall be Qualcomm’s most aggressive entry into the laptop computer SoC market thus far, with Qualcomm dedicating much more in the best way of engineering assets – and advertising and marketing assets – into creating their newest era Arm SoC for Home windows laptops. Backed by Qualcomm’s customized Arm CPU core, Oryon, the corporate is aiming to make the Snapdragon X Elite a watershed second for the Snapdragon model, each carving out a chunk of the profitable Home windows laptop computer market whereas additionally setting the stage for a era of much more highly effective smartphone SoCs in 2024.

As outlined within the firm’s keynote and related announcements finally week’s Snapdragon Summit, Qualcomm is extremely happy with (and gung-ho on) the Snapdragon X Elite, and they’re keen to point out off what it could possibly do. With benchmark scores that, in some circumstances, are beating a number of the greatest chips from Intel, AMD, and Arm, Qualcomm believes they’ve a sizzling hand with the Snapdragon X Elite, and so they need to present it off as quickly as attainable. Even even though precise retail units received’t be accessible till mid-2024 – at the very least one other 6 months from now.

As a part of that promotional push, alongside the general public disclosures finally week’s occasion, Qualcomm additionally held an embargoed occasion for press and analysts to additional show the Snapdragon X Elite and its efficiency. In a closed-door benchmarking session, the corporate set out a collection of Snapdragon X Elite reference design laptops pre-loaded with common benchmark software program, most of which the corporate had additionally utilized in its efficiency claims from final week’s keynote periods. The aim of the session was to show in entrance of the gathered crowd that Qualcomm’s reported benchmark scores have been no exaggeration or anomaly, and that the corporate had units up and operating proper now that might hit these chart-topping scores. Because the saying goes, “seeing is believing,” and Qualcomm’s benchmarking session was all about ensuring the Snapdragon X Elite was being seen in motion.

With units not arriving till subsequent yr, in the present day’s disclosure may be very a lot a preview. And a hands-off one at that; whereas the press was allowed to watch the benchmarks in motion, we weren’t allowed to the touch the laptops (least we attempt to run CPU-Z or Crysis on them), as Qualcomm is holding again on disclosing a number of the technical particulars of Oryon and the Elite SoC altogether. So whereas we are able to report on what we noticed, it’s vital to notice that these are Qualcomm’s benchmark numbers, and never our personal. Although with Qualcomm choosing extremely standardized benchmarks, they’re attempting to maintain issues as comparable and above-board as attainable.

Aggressive Context: Timing Is Every little thing

A part of Qualcomm displaying off so many benchmark outcomes so quickly is, merely put, a matter of pleasure. It’s not day-after-day your organization will get to point out off a brand new SoC anchored by a customized CPU core developed by a crew of engineers, that, by and enormous, attracts from the CPU architects behind Apple’s extremely performant A/M collection SoCs, by no means thoughts their collective expertise from even earlier tasks. The establishment of the Nuvia team virtually 4 years in the past made waves on the time because of the stacked expertise and the potential to disrupt the present CPU core ecosystem, and whereas the Qualcomm acquisition modified that trajectory from servers to cellular units, the potential for disruption remains to be there.

If nothing else, the introduction of a clear sheet high-performance CPU design makes Oryon essentially the most fascinating story in {hardware} tech proper now, and essentially the most fascinating factor to occur within the CPU area because the introduction of Apple’s M1. And if the Snapdragon X Elite as an entire can ship on these efficiency claims – CPU, GPU, and NPU – then that makes issues all of the extra fascinating. In consequence, Qualcomm may be very, very desperate to show the {hardware} that going to be behind their next-generation laptop computer, handset, and automotive designs.

On the identical time, nevertheless, revealing Snapdragon X Elite and publishing benchmark figures for it’s all about timing. Qualcomm desires to manage the narrative as a lot as attainable – they need the Snapdragon X Elite to be seen now and in comparison with rival {hardware} now. Not next day, subsequent week, or next month.

In the end, the 6+ month hole till retail units launch implies that the competitors for Qualcomm’s upcoming SoC isn’t going to be in the present day’s chips such because the Apple M2 collection or Intel’s varied flavors of Alder/Raptor Lake. Nearly everybody goes to have time to roll out a brand new era of chips between every now and then. So whereas Qualcomm’s SoC could also be prepared proper now, we’ve but to see what they’ll be competing in opposition to in premium units. That doesn’t make in the present day’s benchmark disclosure any much less enlightening, but it surely implies that Qualcomm is aiming at a transferring goal – beating Apple, AMD, or Intel in the present day just isn’t a assure that it’ll nonetheless be the case in 6 months.

In any case, apart from having the ability to decide their competitors, an early preview additionally offers Qualcomm an opportunity to line up OEM companions for his or her new SoC. The 8cx chips, three generations in all, have been successively unpopular. Even though the software program aspect of the Home windows-on-Arm ecosystem lastly got here collectively within the final couple of years with Home windows 11, the 8cx Gen 3 has been the least-adopted chip from an OEM standpoint; solely a handful of units, such because the Lenovo ThinkPad X13s and Microsoft’s not-a-Snapdragon Floor Professional 9, have been launched primarily based on it. So Qualcomm want to reverse that trajectory by lining up extra OEMs and extra designs for what is meant to be a a lot higher-profile chip launch for the corporate.

The Reference Laptops: Massive & Smaller

Diving proper into issues, for his or her benchmark demonstration, Qualcomm put collectively two reference design laptops. Dubbed “demo config A” and “demo config B”, the programs are primarily DTR-style and skinny & mild laptops, considerably within the vary of Apple’s 14-inch and 16-inch MacBook Professional designs. Qualcomm received’t be promoting laptops primarily based on these designs, however they’re meant to point out off what OEMs can do with the {hardware}, and the way the {hardware} would carry out in comparable designs.

“Demo Config A”, a high-performance 16.5-inch DTR reference laptop computer design

Showcasing reference designs for efficiency functions may be very typical general for Qualcomm. That is additionally how the corporate handles their smartphone testing – in most years we’ll get to benchmark a Snapdragon cellular SoC operating in a Qualcomm Reference Gadget (QRD) as a preview earlier than retail units can be found. Though they’re not a vertically built-in firm, they nonetheless have to develop units to check their very own {hardware}, and so they like to make use of these to initially showcase their designs.

Moreover differing in bodily dimension, the 2 laptop computer designs have totally different efficiency and TDP envelopes. The excessive efficiency (A) system will get an equally excessive efficiency Snapdragon X Elite chip, with a most clockspeed of 4.3GHz for two core turbo, and three.8GHz for all-core turbo. The smaller skinny and light-weight system (B) is clocked decrease, at 4.0GHz for two core turbo and three.4GHz for all-core turbo. The laptops have decrease TDPs as properly; Qualcomm is disclosing what they name “system TDPs” for the laptops, that are 80W and 23W respectively. This places the skinny and light-weight laptop computer on the sting of having the ability to be passively (fanless) cooled, however right here Qualcomm is utilizing energetic cooling on each laptops.

Notably, right here, each laptops are paired with LPDDR5X reminiscence operating at 8533Mbps/pin. Whereas different laptop computer distributors additionally help LPDDR5(X), none of them help these knowledge charges with their present chips (most of which high out at LPDDR5X-6400). So the Snapdragon Elite X programs are getting a roughly 33% reminiscence bandwidth bonus proper off the bat – one thing that’s particularly useful in GPU benchmarks.

“Demo Config B”, a conveyable, 14.5-inch reference laptop computer design

In any other case there’s little to say concerning the programs, particularly as we didn’t get the possibility to the touch them. They’re primarily meant to present us a take a look at the Snapdragon X Elite’s efficiency in TDP-constrained and unconstrained eventualities, showcasing how efficiency nonetheless stays excessive even whereas cutting down in energy.

The Numbers: Noticed vs. Qualcomm Official Steering

In addition to offering the laptops, Qualcomm additionally supplied official steering numbers for the {hardware}, mirroring the outcomes that have been of their keynote slides from earlier days. The aim of their official numbers was to reveal the official efficiency vary for his or her {hardware}, with the demo models available to show that they’re truly attaining these numbers. In different phrases, the official steering is the efficiency Qualcomm has been getting within the lab, whereas the press was there to watch that they have been attaining these numbers with real-world {hardware}.

As an apart, I’m instructed that the majority of Qualcomm’s official knowledge was put collectively by former AnandTech guru Andrei Frumusanu. Which helps to clarify why Qualcomm’s official figures are primarily based on averaging out three runs of the varied benchmarks, with the outcomes offered as ranges. It’s a degree of statistical finesse that few (if any) producers present.

For his or her demonstration, Qualcomm lined up Cinebench, Geekbench, UL’s PCMark 10 Functions, Procyon AI, and 3DMark Wildlife Excessive benchmarks, in addition to Kishonti’s GFXBench 5.0 Aztec Ruins. At first blush, it’s an uncommon mixture of benchmarks for a laptop computer demonstration, as many of those benchmarks are higher recognized for being utilized in smartphone and pill evaluations. So it’s not particularly a PC-centric set of benchmarks.

However with Qualcomm wanting to place their greatest foot ahead – partly by making certain all the things run right here was with native Arm binaries – it’s not shocking to see them sticking to a core set of benchmarks which are as equally optimized and examined for Arm as they’re x86. In accordance with Qualcomm, the benchmark determination was primarily based on, partly, utilizing benchmarks they steadily encountered in laptop computer evaluations. In both case, it implies that we’re taking a look at a slate of Arm-native benchmarks, with none x86 emulation at hand.

Total, the entire outcomes we noticed from the reference laptops fell inside Qualcomm’s official efficiency ranges – if not exceeding them barely within the case of PCMark 10 Functions. Whereas Qualcomm’s alternative of benchmarks could or will not be cherry-picked to an extent, the numbers they’re reporting are correct.

I’ve additionally put asterisks on a couple of benchmarks, although for various causes. For the Procyon AI benchmark, the Qualcomm units have been operating in opposition to Qualcomm’s personal Snapdragon Neural Processing Engine (SNPE) runtime, which allowed them to entry the NPU features of the SoC. So this was not executed utilizing a vendor-neutral API, similar to DirectML. Moreover, in Qualcomm’s keynote slides, they have been evaluating knowledge to AMD and Intel chips operating CPU paths; so the tip result’s an NPU vs. CPU comparability, a really apples-to-oranges setup.

As for the Linux Geekbench outcomes, as a result of Qualcomm doesn’t but have fan management working below Linux, these programs have been operating with their followers on full blast. Whereas the Home windows programs have been operating with extra typical fan ramp curves, and thus didn’t benefit from the Linux laptops’ successfully limitless thermal setting. Regardless, the first goal of the Linux demo was to showcase that Linux was engaged on the Snapdragon Elite X as properly – that it’s not only for Home windows – as Qualcomm has goals of getting the SoC into Linux laptops as properly.

Placing the Numbers in Context: Snapdragon X Elite vs. The World

The validity of Qualcomm’s official numbers apart, they imply little or no on their very own. We’d like some form of efficiency comparability to present them some context for simply how properly Snapdragon X Elite is performing.

Sadly, Qualcomm’s eclectic alternative of benchmarks makes any form of third-party efficiency comparability very tough. AnandTech itself doesn’t evaluate a lot in the best way of laptops, and even throughout our writer’s whole footprint, our sister websites like Tom’s Hardware, Laptop Mag, Tom’s Guide, and TechRadar don’t use most of those benchmarks in laptop computer evaluations. By no means thoughts the truth that some, like Cinebench 2024, are barely two months previous.

In consequence, we don’t have enough (and constant) knowledge of our personal to make use of in a efficiency comparability. So what we’re left with are Qualcomm’s personal efficiency slides, which current knowledge from choose Intel, AMD, and Apple programs.

Particularly, Qualcomm has collected knowledge from a 2023 Razer Blade 15 (Core i7-13800H) operating at an unconstrained TDP, a 2023 Asus ROG Zephyrus G14 (Ryzen 9 7940HS) operating at an 80W energy restrict, and a 2022 13-inch MacBook Professional (M2). Notably, this group doesn’t embrace Intel’s high cellular chip, the Core i9-13900H, but it surely does cowl AMD’s high chip. In the meantime the M2-powered MBP13 is the least-constrained possibility for displaying off the vanilla M2 chip, but it surely lacks the larger CPU and GPU core counts of the Professional/Max variants used within the 14-inch and 16-inch MacBook Professional fashions.

Notably absent are every other Qualcomm-powered notebooks. So we don’t have any Snapdragon 8cx Gen 3 knowledge to make use of as a comparability.

Qualcomm was fast to point out off their Cinebench 2024 leads to their keynote handle, and understandably so given how common of a CPU benchmark it’s. Pulling forward in single-threaded efficiency isn’t any small activity, given the constraints of each extracting extra IPC from a given workload, and the dearth of headroom to enhance clockspeeds as distributors as soon as might.

Multi-threaded outcomes are a bit extra lopsided; many of the different chips on this chart solely have 8 CPU cores. Even within the case of the 13800H, solely 6 cores are efficiency cores (and the opposite 8 effectivity cores), which in a take a look at of pure CPU throughput would put the Intel system on the again foot.

Geekbench 6.2 is a narrower cluster of outcomes.  After all, Qualcomm has themselves profitable in single-threaded efficiency twice-over (each SDX Elite programs), although as Geekbench is a composite rating of a number of sub-benchmarks to start with, it’s arduous to say how the chips examine in any given activity. Issues open up with multi-threaded outcomes, as is predicted, although once more, not by as a lot as Cinebench.

Switching over to GPU benchmarks, now we have GFXBench Aztec Ruins. That is arguably the oddest alternative of inclusion for a laptop computer benchmark suite, as, particularly in “regular” mode, Aztec Ruins is designed for smartphone benchmarking. By no means thoughts the truth that this benchmark is 5 years previous.

Qualcomm’s outcomes have them properly within the lead right here. However given how unconventional this benchmark is within the PC area, I’ve vital doubts over whether or not it’s been meaningfully optimized for inside Intel and AMD’s driver units. The M2 is probably going the one significant comparability.

3DMark Wildlife Excessive, in contrast, is a extra of a related GPU benchmark. Nevertheless it’s nonetheless a cross-platform benchmark written with smartphones in thoughts – on this case utilizing the extra strenuous “excessive” setting. Usually, the graphics rating of this benchmark is reported, reasonably than the body fee. So it’s not a pure GPU benchmark.

In any case, Qualcomm has in recent times gone in heavy with their built-in GPU designs within the smartphone area – normally outpacing the latest Apple cellular SoC – so it’s not too shocking to see proof of that right here as properly. Something on the extent of the M2 ought to have little hassle outperforming present x86 iGPUs, although I’m a bit stunned that Qualcomm isn’t pulling additional forward of the M2 as properly.

Regardless, being that these are Home windows PCs, I’m extra eagerly awaiting benchmarks utilizing Home windows video games. Whereas Qualcomm must pay the x86 translation penalty, actual video games are going to be far more related in the direction of real-world efficiency than an artificial benchmark is.

Lastly, for the sake of completeness, I’m additionally together with Qualcomm’s final slide for the UL Procyon AI benchmark. As famous earlier, these outcomes are extremely deceptive with out extra context: solely the Qualcomm chips are operating on an accelerated (NPU) backend. The Intel and AMD chips are utilizing the generic Home windows ML CPU backend.

With that mentioned, this case just isn’t Qualcomm’s fault – the Intel chip doesn’t have an equal NPU, and AMD’s isn’t sufficiently uncovered for use on this benchmark – so their outcomes aren’t incorrect. However that is, at greatest, an aspirational benchmark, displaying the advantages of getting an NPU versus doing issues on the CPU. Which, by the point the Snapdragon X Elite ships, the entire main distributors shall be delivery {hardware} with NPUs.

Closing Ideas: An Fascinating First Look

Total, Qualcomm’s early benchmark disclosure gives an fascinating first take a look at what to anticipate from their forthcoming laptop computer SoC. Whereas the aggressive efficiency comparisons are poorly-timed on condition that next-generation {hardware} is simply across the nook from most of Qualcomm’s rivals, the truth that we’re speaking concerning the Snapdragon X Elite in the identical breath because the M2 or Raptor Lake is a significant achievement for Qualcomm. Coming from the lackluster Snapdragon 8cx SoCs, which merely couldn’t compete on efficiency, the Snapdragon X Elite is clearly going to be a giant step up in nearly each approach.

In the end there’s nonetheless an important deal we don’t know concerning the Snapdragon X Elite SoC and its Oryon CPU cores, each with regard to its structure and its efficiency (particularly in x86 emulation). However Qualcomm has made sufficient progress up to now that the Snapdragon X Elite warrants maintaining a tally of as the primary units get nearer to their anticipated mid-2024 launch date.