GPU Performance & Power

GPU performance of the new Adreno 640 in the Snapdrago 855 is interesting: The company’s performance claims were relatively conservative as they showcased that the new unit would perform only 20% better than its predecessor. This is a relatively low figure given that Qualcomm also advertises that the new GPU sees a 50% increase in ALU configuration, as well as of course coming on a new 7nm process which should give the SoC a lot of new headroom.

Before discussing the implications in more detail, let’s see the performance numbers in the new GFXBench Aztec benchmarks.

As a reminder, we were only able to test the peak performance of the phone as we didn’t have time for a more thorough sustained performance investigation.

GFXBench Aztec Ruins - High - Vulkan/Metal - Off-screenGFXBench Aztec Ruins - Normal - Vulkan/Metal - Off-screen

Both Aztec High and Normal results fall pretty much in line with Qualcomm’s advertised 20% improvement over the Snapdragon 845. Here the new chipset falls behind Apple’s A11 and A12 chips – although power consumption at peak levels is very different as we’ll see in just a bit.

GFXBench Manhattan 3.1 Off-screen

GFXBench Manhattan 3.1 Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone XS (A12) Warm 7FF 76.51 3.79 20.18 fps/W
iPhone XS (A12) Cold / Peak 7FF 103.83 5.98 17.36 fps/W
Snapdragon 855 QRD 7FF 71.27 4.44 16.05 fps/W
Galaxy S9+ (Snapdragon 845) 10LPP 61.16 5.01 11.99 fps/W
Huawei Mate 20 Pro (Kirin 980) 7FF 54.54 4.57 11.93 fps/W
Galaxy S9 (Exynos 9810) 10LPP 46.04 4.08 11.28 fps/W
Galaxy S8 (Snapdragon 835) 10LPE 38.90 3.79 10.26 fps/W
LeEco Le Pro3 (Snapdragon 821) 14LPP 33.04 4.18 7.90 fps/W
Galaxy S7 (Snapdragon 820) 14LPP 30.98 3.98 7.78 fps/W
Huawei Mate 10 (Kirin 970) 10FF 37.66 6.33 5.94 fps/W
Galaxy S8 (Exynos 8895) 10LPE 42.49 7.35 5.78 fps/W
Galaxy S7 (Exynos 8890) 14LPP 29.41 5.95 4.94 fps/W
Meizu PRO 5 (Exynos 7420) 14LPE 14.45 3.47 4.16 fps/W
Nexus 6P (Snapdragon 810 v2.1) 20Soc 21.94 5.44 4.03 fps/W
Huawei Mate 8 (Kirin 950) 16FF+ 10.37 2.75 3.77 fps/W
Huawei Mate 9 (Kirin 960) 16FFC 32.49 8.63 3.77 fps/W
Huawei P9 (Kirin 955) 16FF+ 10.59 2.98 3.55 fps/W

Switching over to the power efficiency table in 3D workloads, we see Qualcomm take the lead in terms of power efficiency at peak performance, only trailing behind Apple's newest A12. What is most interesting is the fact that the Snapdragon 855’s overall power consumption has gone down compared to the Snapdragon 845 – now at around 4.4W versus the 5W commonly measured in S845 phones.

GFXBench T-Rex 2.7 Off-screen

T-Rex’s performance gains are more limited because the test is more pixel and fill-rate bound. Here Qualcomm made a comment about benchmarks reaching very high framerates as they become increasingly CPU bound, but I’m not sure if that’s actually a problem yet as GFXBench has been traditionally very CPU light.

GFXBench T-Rex Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone XS (A12) Warm 7FF 197.80 3.95 50.07 fps/W
iPhone XS (A12) Cold / Peak 7FF 271.86 6.10 44.56 fps/W
Snapdragon 855 QRD 7FF 167.19 3.83 43.65 fps/W
Galaxy S9+ (Snapdragon 845) 10LPP 150.40 4.42 34.00 fps/W
Galaxy S9 (Exynos 9810) 10LPP 141.91 4.34 32.67 fps/W
Galaxy S8 (Snapdragon 835) 10LPE 108.20 3.45 31.31 fps/W
Huawei Mate 20 Pro (Kirin 980) 7FF 135.75 4.64 29.25 fps/W
LeEco Le Pro3 (Snapdragon 821) 14LPP 94.97 3.91 24.26 fps/W
Galaxy S7 (Snapdragon 820) 14LPP 90.59 4.18 21.67 fps/W
Galaxy S8 (Exynos 8895) 10LPE 121.00 5.86 20.65 fps/W
Galaxy S7 (Exynos 8890) 14LPP 87.00 4.70 18.51 fps/W
Huawei Mate 10 (Kirin 970) 10FF 127.25 7.93 16.04 fps/W
Meizu PRO 5 (Exynos 7420) 14LPE 55.67 3.83 14.54 fps/W
Nexus 6P (Snapdragon 810 v2.1) 20Soc 58.97 4.70 12.54 fps/W
Huawei Mate 8 (Kirin 950) 16FF+ 41.69 3.58 11.64 fps/W
Huawei P9 (Kirin 955) 16FF+ 40.42 3.68 10.98 fps/W
Huawei Mate 9 (Kirin 960) 16FFC 99.16 9.51 10.42 fps/W

Again switching over to the power and efficiency tables, we see that the Snapdragon 855 is posting a ~30% efficiency boost over the Snapdragon 845, all while slightly improving performance.

Overall, I’m very happy with the initial performance and efficiency results of the Snapdragon 855. The S845 was a bit disappointing in some regards because Qualcomm had opted to achieve the higher performance figures by increasing the peak power requirements compared to exemplary thermal characteristics of the Snapdragon 835. The new chip doesn’t quite return to the low power figures of that generation, however it meets it half-way and does represent a notable improvement over the Snapdragon 845.

System Performance - Slightly Underwhelming? Final Thoughts
POST A COMMENT

132 Comments

View All Comments

  • Rudde - Friday, January 18, 2019 - link

    Don't cherry-pick results. Reply
  • Rudde - Friday, January 18, 2019 - link

    What is raw performance? I could calculate som fused multiplies per second for you, but is that 'raw performance'? Reply
  • HStewart - Tuesday, January 15, 2019 - link

    I differ with a lot on this - I think A12/A12X and other ARM related device actually perceive faster because of marketing. Also with App architexture of the OS running on such devices hide the actual performance of chips - I think it specifically depends on what you using the device. For normal word processing, emails and internet - it can easy be shown that same as U series - and this depends on which model - likely dual core x86 and possibly even AMD notebooks - but not a 4+ ghz laptop like my Dell XPS 15 2in1. Keep in mind on a phone and even android tablet or iPad there is less screen to drive. I am talking about real professional software and not apps

    One thing is interesting about 855 design is big core designed - running the primary core at higher speed then other 3 primary core - is smart - this means the primary thread is running at higher speed. I assume that smaller cores would be use threads for background tasks. Intel has a similar designed large single core and 3 minor atom based cores - I would think that device is closer on performance to A12 based devices not the U series.

    My big question is that I think it hard to actual compare performance between any x86 base and any ARM based. It depends on designed of OS and applications running on devices. I would be 100% sure any software that uses AVX 512 would blow any ARM based application with similar abilities. In fact with AVX 512 application it would be big difference with AVX 2 based computer.

    All I am saying is performance depends on application running, not just web browsing and other things
    Reply
  • HStewart - Tuesday, January 15, 2019 - link

    One thing also - the speed of cpu, number of cores, or even node process does not make the performance of device - it how it used with architexture inside that makes the difference. Reply
  • Wilco1 - Tuesday, January 15, 2019 - link

    You're absolutely wrong, A12X can keep up with your beloved laptop - this is the latest and fastest variant: http://browser.geekbench.com/v4/cpu/compare/109702...

    Single-threaded integer score is within 2.5%. Mind you that's a 10W SoC compared with a 65W CPU! I'm awaiting your list of excuses how that is possible...
    Reply
  • goatfajitas - Tuesday, January 15, 2019 - link

    One test on one specific thing. Try 100's Reply
  • Wilco1 - Tuesday, January 15, 2019 - link

    No it's not one test, nor one specific thing. Like most benchmarks there are many different tests and the average is reported. Reply
  • goatfajitas - Tuesday, January 15, 2019 - link

    My point is that in a world of benchmarks, you are looking at it very myopically. ARM isnt anywhere near as fast as x86 in raw power. Very good and super efficient at alot of multimedia tasks though. Reply
  • Wilco1 - Tuesday, January 15, 2019 - link

    What benchmarks??? Name another cross platform benchmark which is NOT a useless browser test. Apart from SPEC, Geekbench is one of the very few benchmarks that allow reasonable cross platform comparisons. Reply
  • goatfajitas - Tuesday, January 15, 2019 - link

    "cross platform" benchmarks are virtually useless. Your grasp of benchmarking in general needs work. It's not apples to apples. Reply

Log in

Don't have an account? Sign up now