Next year looks like it might be very exciting — when it comes to smartphones, anyway. The traditional logic would tell you that Qualcomm’s new Snapdragon 8 Gen 1 chipset will set the standard for flagship Android phones in 2022. But, for the first time in years, that’s not guaranteed. MediaTek is also ready with its own new big-boy chip in the Dimensity 9000, Google’s gone custom, Samsung’s bringing AMD GPUs to its in-house SoCs, and even OPPO is rumored to be working on its own SoC. It’s a smartphone chipset showdown. But will it be a silicon bloodbath, a boring and easy victory for the encumbent, or can customers expect something even better (if less dramatic)? There’s already evidence to be excited in the latest benchmarks.
We don’t usually obsess over benchmarks here at Android Police. While they’re a ballpark indicator of performance, some manufacturers (and even some chipset companies) manipulate them into meaning much less, and they’re just numbers. I’m basically a professional consumer communicating my product experiences to you, when you get right down to it. And years of experience have stressed to me that the holistic whole-phone experience is much greater than the sum of its parts, even if I get excited for things like single-threaded performance gains, too. Measurements for one part in isolation don’t describe a complete product, though. And even if you do like them as a back-of-the-envelope comparison, they may not even measure the right things.
Take it with a grain of salt, but Qualcomm’s engineers are convinced that most machine learning benchmarks right now are inaccurate because they’re doing the wrong kind of math to reflect real machine learning model workloads (calculations with high-precision floating-point numbers rather than low-precision or integer). And that ignores the fact that benchmarks don’t take into account things like power improvements unless you try to measure those separately — and that’s pretty hard to do. Ultimately, looking at each final phone product as a whole is the only complete picture you can get.
Still, the numbers we got for the Snapdragon 8 Gen 1 are worth discussing for one very important reason: Based on two numbers we have to compare, MediaTek’s upcoming chip might actually be winning when it comes to a couple of these arbitrary numbers.
Benchmarking with Anshel Sag (friend of the site and principal analyst at Moor Insights & Strategy)
First, the Snapdragon 8 Gen 1 benchmarks. Some of these scores were a group effort because specific tests crashed on some of the reference devices. (Thanks: Rich Woods, Anshel Sag, and Myriam Joire.)
- Geekbench (5.4.2): 1226 single-core, 3797 multi-core (three-test average)
- Antutu: 991349.66 (three-test average)
- PCMark 16940.66 (three-test average)
- Jetstream 2: 137.559
- Speedometer 2: 120.9 +-0.85
Qualcomm ran these last benchmarks for us after the fact (the 888 and Tensor were tested on my own devices). Some huge AI gains here, especially in integer performance.
In isolation, these numbers are good but not incredible (excluding the AI gains here, which are huge and even Tensor-beating in spots). Some of them barely beat the Snapdragon 888. The place where we were most likely to see performance improvements, based on what Qualcomm told us, is in machine learning and AI workloads, but those benchmarks are a little more complicated.
Qualcomm told us its new chip should be able to break the million-point barrier in Antutu, but that wasn’t true in our testing on average. (Only one of three benchmarks did.) But, surprise! MediaTek says its chip can do the same.
I need to stress, MediaTek didn’t actually let us play with a reference device like Qualcomm did, so we’re comparing benchmarks I’ve proven myself on actual hardware with a company’s claims (read: take them with a massive grain of salt), but MediaTek says its Dimensity 9000 can break a million in Antutu, and it also claims its chip can hit a multi-core Geekbench score of over 4,000, which would actually beat what we’ve seen from Qualcomm out of the 8 Gen 1 in our testing if it’s true.
Again, these are just two numbers that don’t really matter out of a whole pile of numbers that don’t really matter. On top of that, manufacturers will undoubtedly apply their own tuning systems, further muddying the waters. But it is indicative of one significant thing: Even if Qualcomm still ends up leading the pack this year, it doesn’t seem like it will be ahead by a crazy margin like we’re used to. MediaTek is right there with it — or, at least, nipping at its heels. There could yet be some huge gotcha out of any of these chips (more on that in a second), but based on what little we’ve seen so far, the gap is narrower than it’s ever been. And, for customers, it’s a difference that might not actually matter — other specs could be more critical.
These GFXBench numbers didn’t jibe with what we were told to expect — may have been a bug. They weren’t included in the list above, but you can see them here if you’re curious.
Many customers are used to the Android flagship formula and buy phones based on marketed features rather than specs, at least partly because basically all the flagships had the same Qualcomm chip inside them. Customers at the high-end Android postpaid market shop by comparing things like camera performance and what you can do with the phone rather than a spreadsheet of components. (That’s less true in other markets, but we’re primarily concerned with the US.) The spec-based cause of a bigger camera sensor or a newer chip = snore, but the effect-based result of better photos = 👀. If MediaTek (and other custom solutions) can even approach Qualcomm’s level, customers probably won’t directly care which chip is in it, and phones with that hardware will sit right next to Qualcomm-powered models on the carrier store displays.
This is the first time in years that we’ve really anticipated a “flagship” chipset from a company other than Qualcomm — Samsung’s halfhearted Exynos chipsets don’t really count. Moderating that, though, is what I think of as “the rule of firsts.” I’m not sure if it’s a real thing or something I just made up, but it’s like the early adopter rule, or the bleeding edge rule. Basically, the first time you do something you’ve never done before, you’re gonna suck at it. So it follows, the first product in a new lineup or category is almost guaranteed to have issues. And, I should stress, the Dimensity 9000 is MediaTek’s first flagship chipset — they might suck at it.
That’s not an inevitability, though. After all, MediaTek has been making chipsets for years now; this is just the first time it’s gone out of its way to make a really high-end one. And it’s not that the company tried and failed, it’s that it just never tried. And, if it’s not obvious by now, Qualcomm’s pretty good at making high-end chipsets after all the practice it’s had. The Snapdragon 8 Gen 1 is nearly guaranteed to be free of major showstopping issues — even the demo units we played with ran nice and cool to the touch after sessions of intense benchmarking (though scores did slightly fall).
Of course, there have been some stumbles, as anyone that owned a smartphone with a Snapdragon 808 and 810 can easily remember. Even this last generation of phones and the Snapdragon 888 saw some variety in performance across devices (and, presumably, cooling solutions). Now that MediaTek is ready to move upmarket, although I don’t expect it to, Qualcomm really can’t afford to stumble. With supply chains already stretched to their limits for consumer gadgets and companies reengineering products to fit what’s available, a misstep now could seriously cost Qualcomm — if not immediately in 2022, then later in 2023 if engineers designing phones today decide that things have changed.
Speaking to analysts more recently, I’m also told that the lack of mmWave 5G in the Dimensity 9000 could be less of an impediment in the US now, excluding Verizon. With the rise of better low- and mid-band 5G, only big red is leaning on its impractical mmWave 5G crutch and associated marketing. In the US postpaid market (where flagships live), it’s the carriers that sell phones, and T-Mobile and AT&T could easily entertain the Dimensity 9000 and its lack of mmWave. At the Snapdragon Summit, a T-Mobile executive straight-up told me yes when I asked if the company would be willing to sell a hypothetical $1,000+ flagship device with a Dimensity 9000 and no mmWave 5G. That’s no guarantee it will happen, but it shows a willingness.
MediaTek has also seen some “big wins” (as the corporate-speak goes) in the last few years, and the momentum is behind it. The new e-ink Kindle and Peloton’s hardware are both powered by MediaTek chipsets, and they’re sure to be big movers. And, as OnePlus continues its stratospheric rise, it’s a journey made partly on the wings of MediaTek and the Dimensity 1200 inside the Nord 2. For all we know, a company like OnePlus could already have a flagship bound for our shores with a Dimensity 9000 in it. It’s a move that would hardly make MediaTek a household name, but, to be fair, neither is Qualcomm, and on carrier shelves it wouldn’t matter.
Customer priorities are also shifting amid the rise of foldables and the expectation that phones will last longer and longer as customers keep them longer — and, importantly, keep getting updates. Chipset features that better work with new form factors and software support could be equally important, and those are areas where Qualcomm might have an edge. On top of that, AI performance improvements for applications like ambient computing are where the real gains are to be had now (and I suspect Qualcomm will soundly beat MediaTek there).
But whether Qualcomm or MediaTek has the best chip (my money, for the record, is on Qualcomm given its experience), as long as they’re even remotely close to one another, it’s a big win for customers — and that’s ignoring in-house custom designs from companies like Google and OPPO, or the possibility that Samsung’s Exynos might actually be good again. We may finally see chipset competition at the high-end in the US market for the first time in nearly a decade.
In 2021 we started to see prices inch down after they flew through the roof in 2020. More competition in 2022 could drive down chipset prices and further reduce product prices, maybe making a dent in the face of inflation or even supply chain constraints. And with the new types of specialization that companies like Qualcomm and Google are putting in their chipsets when it comes to specific workflows (Qualcomm’s always-on camera and Leica camera processing, Google’s Tensor-optimized real-time translation and photo effects, etc.), a return to chipset competition could push even more variety in features in response.
This is going to be an exciting year for smartphones — and, with supply chain constraints, one that could see a lasting impact for whichever company “wins.” But one thing is for sure: With the extra competition, every customer this year is a winner, too.
Maybe someday you’ll drive a car that supports it
About The Author