First Snapdragon 850 Benchmarks Disappoint (Updated)

Posted on August 15, 2018 by Paul Thurrott in Hardware, Mobile with 29 Comments

Updated: I’ve heard from two credible sources close to Qualcomm that this benchmark does not represent the actual performance of the Snapdragon 850. –Paul

Early Snapdragon 850 benchmarks suggest that Qualcomm’s first PC-centric chipset may not bring hoped-for performance gains.

In a benchmark posted to Geekbench and first spotted by WinFuture.de, a Snapdragon 850-based PC scored 2263 in a single core test, a 25 percent improvement over the Snapdragon 835. But the 850-based PC netted only a 7.3 percent gain in a multi-core test, with a score of just 6947, vs. 6475 for the 835.

Qualcomm, you may recall, claimed that the Snapdragon 850 was optimized for performance and should see a 30 percent gain, overall, when compared to the 835. That’s true of both the processor, which runs at a higher clock rate, and the graphics processor.

So. What are we to make of this?

Well, it’s only one test. And Qualcomm, of course, has committed to a long-term strategy of designing custom chipsets for each market it targets. We’ve seen rumors about a coming Snapdragon 1000 as well, and that one may be aimed at desktop PCs.

But this is not good news, for sure. Today’s Snapdragon 835-based PCs are basically a non-starter. And we need to see bigger gains than this.

Join the discussion!

BECOME A THURROTT MEMBER:

Don't have a login but want to join the conversation? Become a Thurrott Premium or Basic User to participate

Register
Comments (34)

34 responses to “First Snapdragon 850 Benchmarks Disappoint (Updated)”

  1. Rcandelori

    Not sure how a 25% improvement in single-core performance is "disappointing". When was the last time we saw such a leap in performance from an intel mobile chipset? They only got their claimed 40% improvements this year by going quad core everywhere.

  2. wright_is

    The problem is, benchmarks are synthetic and do not necessarily show how the percieved performance has improved.

    These sorts of processors, like low end Intel processors, aren't designed to be speed demons and complete benchmarks in record, or fast, times. What is important is the user experience doing common tasks.

    Do applications open more quickly? Do the applications run smoothly? Do web pages open and react more quickly? Are they "smooth enough"? Those are the criteria that will really show whether the processor has been significantly improved.

    Geekbench pushes the processor and, according to them, "simulates" real world use. But it avoids straightforward problems and doesn't do repetitive tasks, which is what a typical user will do. So it will be interesting to see how it actually performs under real conditions.

  3. PeterC

    Is this why andromeda didn’t make the cut?


    shame really, looks like Apple is going to have a significant lead in its own ARM iOS based product categories (Mac OS / iOS arm laptop stuff). clearly Qualcomm will improve but it looks to me like Apple will gain control and advantage in this new area whilst win 10 arm lags behind.


  4. skane2600

    An increase in performance is great but the 850's competition isn't the 835. What matters to the market is how it compares to offerings by Intel and AMD.

  5. PeteB

    Welp, so much for the "Windows on ARM" nonsense.


    Seems Microsoft has stopped talking about it too.

  6. glenn8878

    A second generation product. I guess it takes time to ramp up, which is what Intel has been doing for the last decade. These are phone substitutes. Not fast enough as PCs and even as decent tablets since Windows has a performance penalty. So it’s whether you rather use a larger screen when not using your phone. I rather buy a large screen television and Roku for less.

  7. MikeCerm

    This looks like good news to me. The processor easily beats the Pentium in the Surface Go in both single- and multi-core performance, while (presumably) consuming far less power. So maybe instead of 5-6 hours of battery life, an ARM-based Surface Go could hit 8-10 hours? Sign me up!


    Only problem is that Qualcomm would have to drop the price pretty drastically to compete with Intel, and I don't know that they care to do that.

  8. Silversee

    As ever, these PC benchmarks are somewhat meaningless if they are run using the x86 binaries. Windows benchmark suites need to be recompiled for ARM native code for the Snapdragon. Only this will tell you what the real performance potential of this platform is.

    • skane2600

      In reply to Silversee:

      Performance potential, yes, market potential, not necessarily. You need popular programs to be available for the platform and so far there's not much interest in modifying and compiling x86 programs for ARM.

      • Silversee

        In reply to skane2600:

        That will change in due course. ARM is too important for Microsoft to walk away from, and eventually developers will catch on.

        • skane2600

          In reply to Silversee:

          I think you've made two big assumptions. In recent years there's been a kind of ARM frenzy but it doesn't seem to based on anything that substantial. Lower power is very important for smartphones, but of lesser importance for PCs.


          Keep in mind that developers never caught on to Windows Phone and Microsoft eventually walked away from that market. So there's a precedent for MS's lack of patience and developers reluctance to develop for a platform without a clear business case.

  9. Subhadip Sen

    Given it's just a lightly overclocked 845, this is not surprising. Perhaps the next-gen (Snapdragon 1000?) will see a true x86-class CPU.

  10. HellcatM

    Isn't this chip suppose to be optimized for Windows as opposed to the 835 which isn't so even though the speed isn't faster it should run Windows better right?

  11. SRLRacing

    Seeing as most of those every day tasks an arm based PC will largely be used for are heavily single threaded I think the real world performance should be much improved.

    • wright_is

      In reply to SRLRacing:

      Plus the bechmarks are designed to push the processor to its computing limits, but not to see how well it holds up doing "typical" activities. It is like saying a Ford Mondeo is slower than a Ferrari, yet you try getting the monthly shop into the back of a Ferrari! ;-)

  12. FalseAgent

    wait a minute...a 25% gain in single core performance is actually quite a big deal, because that's where the 835 was lacking. The 835 was a already (relatively) competent multi-core performer. Adreno also beats Intel's rubbish iGPUs.

    • F4IL

      In reply to FalseAgent:

      You are correct, but the context here is that it doesn't look good for windows (an OS written for x86_64 with decades of legacy and evolutionary baggage).


      Following this trajectory, WOA won't be a realistic alternative before 2020.

      • FalseAgent

        In reply to F4IL:

        no, Windows itself runs fine. Windows has been completely ported to ARM and all of it pretty much runs native, including whatever baggage it has. Windows RT (Windows 8) ran on the Nvidia Tegra 3 (which is a far slower CPU), and that was basically WOA as well. It just didn't have x86 emulation.


        The thing slowing WOA down now is the emulation of x86 apps. WOA has already happened, but native ARM64 exe's hasn't.

        • skane2600

          In reply to FalseAgent:

          It's not clear that Windows has been "completely" ported to ARM. I'm not aware of any definitive statement from Microsoft claiming 100% compatibility between their x86/x64 implementation of Windows and their ARM implementation. I wouldn't be surprised if some of the more obscure APIs didn't make the cut.

  13. tommorton14

    How can Qualcomm be so far behind the Apple A11 Bionic ? The A12 will be released along side this chip? If Apple adds track-pad and mouse support to the Ipad Pro Windows on Arm is DOA!

    • Subhadip Sen

      In reply to tommorton14:

      Simple, Apple has a shitload of money. They can pour billions into R&D Which Qualcomm simply cannot afford. The A11 Bionic has much larger cores, and is overall a larger chip, and hence costs much more to manufacture. The average iPhone shipping with A11 Bionic is somewhere in the $700-$800 range, so they can easily afford to make the largest die possible, while Qualcomm has to consider economics. The Snapdragon 845 can be found in devices shipping for $400-$500. You may think they have a monopoly, but they don't. If they screw up, their two largest customers (Samsung and Huawei) will ditch them. (Samsung rejected Snapdragon 810, and pretty much uses Exynos everywhere outside of US and China; while Huawei has all but transitioned their entire lineup to their own Kirin SoCs.) Xiaomi (3rd largest) is also developing their own SoC.


      Also, the Geekbench heavily favours Apple's architecture. Other benchmarks and real world applications paint a different picture. Yes, Apple still leads, but not by quite as much much. And in GPU and some multi-threaded workloads, the 845 is faster. That said, A12 will make another leap forward, surely. But there are rumours Qualcomm will respond with their own premium SoC, Snapdragon 1000, early next year. Let's see.

  14. jimchamplin

    Sounds like they need to work on their SMP implementation.

  15. roastedwookie

    :))) Well, it looks like MS's grand WinARM project will be the same crippled crap like it is now with the 835 SOC. But hey, they expect people to take the bait and spend $999 + taxes, 1000-1200 EUR on such a PC :))) Or, at least HP does

Leave a Reply