The Snapdragon 855 is a benchmarking beast in a year that needs the power

Qualcomm Snapdragon 855 testing
Qualcomm Snapdragon 855 testing (Image credit: Daniel Bader / Android Central)

Last year, we told you that while the Snapdragon 845 was an incredibly powerful chip that pushed Qualcomm forward in a number of ways, its benchmark numbers didn't really matter. At least they didn't back in February, before the first devices running the new flagship processor were on the market.

See, every year, prior to releasing its biggest-and-best platforms into the wild, Qualcomm builds a bunch of bulky, nondescript Android devices that manufacturers and, eventually, press can handle for a couple of hours in order to ascertain the inevitable numbers bump for that particular year. On their own, the numbers are pretty meaningless — what does a score of 360229 in AnTuTu 7.1.1 mean in isolation, anyway — but looked at in retrospect they tell a pretty interesting story.

Because now we have the benefit of hindsight, and I'm able to look back at those reference devices and see how production products, like the Pixel 3 I used alongside the Snapdragon 855 RD, compared to Qualcomm's own Snapdragon 845 equivalent. And the results are quite interesting, mainly because, while the Pixel 3 is a little slower in almost every test, it's within the margin of error (around 5%).

But the Pixel 3 is a fully-formed product, with working radios and a compact design and software that's been highly optimized. And it's slower than a reference device Qualcomm cobbled together with spare parts and a prayer? What does that say about the numbers I'm about to show you in relation to products like the Galaxy S10 or Pixel 4 or whatever else is released running the Snapdragon 855 in 2019? I'm not really sure, other than to reinforce that you should take these numbers with a hefty grain of salt and not read too much into them.

With that being said, let's talk numbers. First, a couple things:

  • All tests were done three times and the result was the average of those three tests.
  • Despite being offered the chance, we didn't perform any A.I.-specific tests on the Snapdragon 855 reference device, as Qualcomm says most have not been optimized for its Hexagon DSP and may not reflect the significant advancements in this year's processing abilities over the S845.
  • These tests were done over a 90-minute timeframe on Sunday, January 6, 2019.
  • Happy new year!
Swipe to scroll horizontally
BenchmarkQualcomm Snapdragon 855 RDGoogle Pixel 3% difference
AnTuTu 7.1.1Row 0 - Cell 1 Row 0 - Cell 2 Row 0 - Cell 3
Overall35934623634452%
CPU1184586810374%
GPU15690310952843%
UX734674745155%
Memory1051911372-7.5%
GeekbenchRow 6 - Cell 1 Row 6 - Cell 2 Row 6 - Cell 3
Single core3501234849%
Multicore11189789642%
GFXBench 4.0Row 9 - Cell 1 Row 9 - Cell 2 Row 9 - Cell 3
ES 3.1 Manhattan 1080 offscreen714461%
*ES 3.0 Manhattan 1080 offscreen1026362%
*ES 2.0 T-Rex 1080 offscreen61610%
*ES 3.1 Car Chase 1080 offscreen423520%
Jetstream 1.1115.5369.41354%
Speedometer11837216%

Just glancing at the numbers, it's clear that the Snapdragon 855 PD performs traditional CPU, GPU and Javascript benchmarks very well — between 20% and 75% better, on average. Some, like the Javascript-heavy Speedometer browser test, are huge jumps from the 2018 flagship; others, like the CPU-bound Geekbench and AnTuTu, are a more reasonable 50%.

Except that a 50% year-over-year jump is massive in this space. Part of the increase comes from Qualcomm's new Prime Core Kryo 485 layout, which offers a single high-performance core additional clock speed when needed, along with double the L2 cache, compared to the Kryo 385 in the Snapdragon 845. The new cores are also based on ARM's Cortex-A76 architecture, a bump over the A75 on which the 385 was based. Then there's faster LPDDR4 memory and a host of other minor improvements that lead to a generational performance leap we haven't seen since before the Snapdragon 800.

This is the biggest year over year leap we've seen since the Snapdragon 800, and that's just counting CPU and GPU performance.

On the Adreno 640 GPU side, Qualcomm only claims a 20% performance boost over last year's 630, but these benchmarks show more variation than that. Still, it's extremely fast and few games will likely give it any trouble.

What's also not being shown here is power conservation: Qualcomm claims that despite these huge gains, the Snapdragon 855 is considerably more efficient than its predecessor, largely thanks to its new, more efficient 7nm manufacturing process.

Then there's the stuff that traditional benchmarks don't cover. Qualcomm spent a lot of time — most of the time during our briefing, in fact — emphasizing that the above benchmarks offer a tiny proportion of the total potential of this chip, which is optimized for AI and machine learning tasks that most Android developers have yet to incorporate into their applications. Those tasks, according to the company, should be performed at speeds three or four times faster on the Snapdragon 855 than the 845.

All of these advancements come at a time when Qualcomm's competition is firing on all proverbial cylinders. In late 2018, Huawei announced its Kirin 980 SoC with many of the same claims as we see above, just a few months earlier. And while the Snapdragon 855 should trounce the Kirin 980 in most synthetic benchmarks, Huawei seems to be a generation ahead by many metrics, and is being held at bay primarily by Qualcomm's outsized U.S. market share.

Finally, there's Apple's A12 Bionic, which was announced alongside the iPhone XS in September. I did a few cursory benchmarks to see, given the Snapdragon 855's considerable advancements in 2019, whether Apple's internally-designed SoC was still the industry winner. And, well, see for yourself.

Swipe to scroll horizontally
BenchmarkQualcomm Snapdragon 855 RDiPhone XS% difference
AnTuTu 7.1.1Row 0 - Cell 1 Row 0 - Cell 2 Row 0 - Cell 3
Overall3593463354957%
CPU118458127126-6.82%
GPU15690313115320%
UX734676480613%
Memory1051964806-15%
GeekbenchRow 6 - Cell 1 Row 6 - Cell 2 Row 6 - Cell 3
Single core35014811-27%
Multicore1118911483-2.56%
Jetstream 1.1115.53269.7-57.16%
Speedometer118242-51.43%

I was expecting worse — a lot worse. Apple's had a pretty sizeable lead in almost every metric over the flagship Snapdragon 8xx SoC over the past few years, and while you can certainly say that's still true, this is the closest they've ever been. Apple still owns Qualcomm on Javascript optimization — heck, it beats Intel — and its GPU is still considerably faster, but the overall picture looks pretty good for Qualcomm right now, and that's a feat unto itself.

I'll leave it there for now. There's no point getting too hung up on these numbers until the Galaxy S10 is released in March and we had a shipping phone with a commercial version of the Snapdragon 855 running retail Android software. But I'm also really hopeful that, alongside the proliferation of 5G and the maturation of AI-based applications, this'll be a year to remember for performance in general in the Android space.

Daniel Bader was a former Android Central Editor-in-Chief and Executive Editor for iMore and Windows Central. 

23 Comments
  • Where's the comparison between Kirin 980 vs SD 855?
  • The Pixel 3 is not a valid comparison in my opinion, since Google had the wonderful idea of capping the SD845s max frequency. With an uncapped CPU the difference would be smaller. Still impressive numbers, tho!
  • Sorry but other media outlets and testing show that the performance gains really are not much to talk about. The performance, while perfectly fine, is actually a little underwhelming.
  • Everyone has to go out and buy a new $1,000 phone now as their current phone will suddenly perform like a phone from 10 years ago.
  • Luckily, 1000 dollar phones will not be the only ones with the 955!
  • I'd like to see the SD835 compared as well for fun (and because I still have an S8)
  • Still using a Snapdragon 800 with modded software. Still runs decently. I can never see justification to buy the new SD emabled phones for a slight boost.
  • Running 810 here and absolutely no complaints
  • Your iPhone XS "Memory" figure in the chart is incorrect. :-)
  • Hey, if Apple's phone sales keep declining, they can always expand their chip sales to Android vendors. Then we won't constantly hear how Android's CPU/GPU performance lags behind Apple.
  • In all honesty , these numbers are disappointing. It doesn’t compete across the board with last years iPhone XS. It seems to be the norm now that the best Android phones are a year or more behind Apple in terms of raw performance. You can argue all you want that the end user won’t notice (unless they have 2 devices side by side) However as Android flagships are half the price of the iPhone XS, maybe it’s not so bad !
  • The XR is only 750 retail
  • Kirin 980 is faster in real world usage though and has much better thermal performance.
  • More synthetic benchmarks....99% of users would never notice the difference between the two. Email, Twitter, Instagram etc all run fine on $250 android phones now. Spending $1000 is just wanting to be apart of the “cool” crowd.
  • If you needed any more proof that people don't care about raw power in a void just look at Apple's underwhelming sales of the iPhone XR: that whole phone can basically be summed up as "all the power of the flagship iPhone with corners cut elsewhere to bring down the cost," and it seems like people aren't impressed. Of course higher processing performance is a good thing, and I'm super psyched to see improvements in that area, but there are diminishing returns past a certain point and even the mid-range Qualcomm lineup is past that point, let alone the Snapdragon 945, Kirin 980, and A12 Bionic. Once you get up to those kind of numbers the average consumer just doesn't see the value proposition in taking the extra horsepower over features like display, weight, camera, and style. If it's fast enough, most people simply don't care about whether it's the fastest. All this to say that with prices spiraling ever upward, manufacturers would do well to remember that benchmarks alone don't sell phones: it's important that the device as a whole looks and feels good to use.
  • It's not that. The LCD screen is a different tech than AOS uses for LCD and it covers the entire screen very well. The camera is also tops. Not to mention is at least $250 than the XS!
  • Would anyone be able to help with a link to that wallpaper in the top pic? Thanks
  • I want to see this bad boy overclocked. Smartphone chips should, if they're as high quality and the venerable OMAP 4430, be able to handle 96c before crashing from heat induced instability.
  • I can't think of any reason why the Pixel 3 would underperform the Pixel 3 XL (if anything, it should at least outperform the 3 XL in graphics, driving a lower resolution display), but the numbers indicated for the Pixel 3 seem awfully low. By comparison, my Pixel 3 XL, not rooted and fairly well stocked with apps (only about 5GB free storage) scored the following marks on Antutu 7.1.1: Overall: 279736*
    CPU: 86692
    GPU: 123625
    UX: 54665
    MEM: 14754 *note these are just the scores from the last run. My 'average' overall score was actually marked even higher, at 283807, increasing the variance from the Pixel 3 even further. With such a drastic difference, I'm left wondering if Daniel's phone has either been rooted or perhaps had a lot of apps running in the background - something which is unlikely to have been the case when Qualcomm's reference device was tested. On a separate note, I believe there is a typo where the iPhone XS is shown scoring 64806 on both UX and Memory (presumably the error is the score on memory, when taken in context of the 855 score and % difference calculated).
  • The only number in that article that mattered is 7nm. I want to see better battery life. That's the elephant in the room. After all this time, phones can still barely last a day without recharging.
  • @Wizzy Better battery life comes in small gains from the processor, and not enough to dramatically enhance quality of life - if you can burn through a typical flagship battery in a day now, you'll likely still do so even with processors several generations in the future. The real gain will need to come in the form of a serious evolution of the batteries themselves, either in power storage/volume for simply not needing to recharge, or extreme speed of recharge to minimize inconvenience.
  • « At least they didn't back in February, before the first devices running the new flagship processor were on the market » So they actually did and the first article in February made no sense
  • Looking at this in retrospect, and the real world gains turned out to be insignificant.