GPU Performance & Power

Moving on, it’s time to talk about the GPUs of the systems. The Snapdragon 865’s Adreno 650 is a microarchitectural successor to last year's Adreno 640, increasing the ALUs and ROPs by 50%. Frequency remains the same at 587MHz, and the company promises a 25% performance boost.

The Exynos 990 is more drastic in its GPU changes. Here we see for the first time a chip using Arm’s new Valhall GPU architecture, in the form of the Mali-G77. We’ve discussed the GPU in detail in the deep dive article last year, so be sure to read about the details of the new design there. Samsung LSI employed an 11-core configuration in the new chip, 1 less core than last year’s G76MP12. This is compensated by clocking the design higher at up to 800MHz, up from 702MHz. The higher clock speed does however cost some additional voltage to reach, with the Exynos 990 now peaking at 712mV compared to the 662mV of the previous iteration, although both designs should be clearly operating at lower than nominal voltages of the process nodes.

Beyond the new GPU hardware, it’s also important to note the new chips are the first of their kind to support LPDDR5, which should bring some good efficiency upgrades to bandwidth hungry tasks such as 3D rendering on a GPU.

3DMark Sling Shot 3.1 Extreme Unlimited - Physics

Starting off with 3DMark Physics, which is actually a CPU test in a GPU thermally constrained scenario, we see both phones doing well. The Exynos 990 here likely schedules things more onto the A76 cores, and that’s why performance is less than that of the Snapdragon 865 which here takes the leadership position in the benchmark. Throttling isn’t very prevalent on either device, but for some reason the Exynos Ultra device throttled more than the S20+.

3DMark Sling Shot 3.1 Extreme Unlimited - Graphics

Moving onto the graphics subtest, we’re seeing an extremely stark contrast in scores. The Exynos 990 is able to keep up with the Snapdragon 865’s peak performance figures, however once throttling kicks in, the scores quickly fall down to more moderate figures. The Exynos S20 Ultra’s performance here is again quite puzzling as to why it’s so much worse than the S20+ – both phones didn’t seem to behave very differently in their thermal behavior, so that’s super weird. The performance deficit here is gigantic, with the phone only sustaining 28% of its peak performance.

Meanwhile the Snapdragon S20 Ultra doesn't throttle here at all, and that is absolutely not normal – this is not a chip that is somehow super-efficient or has amazing cooling. Over the years I’ve encountered a lot of such odd results with Snapdragon phones in this benchmark, but this time around I’ve had enough of the weird behavior and I do think there’s some low-level cheating going on. The phone will actually start heating up a lot more than under other workloads, up to the point that the test will actually crash. I don’t understand how that’s possible that this happens only in one benchmark but not others, and the most logical (and likely) explanation is that there’s some benchmark detection going on. Again, I’ve only ever encountered this issue on Snapdragon phones in this test (and we’re also using a custom APK), so it’s super suspicious, but we’re just short of finding the smoking gun that this is some malicious behavior. In any case, please disregard the results as they’re not representative of real behavior.

A few weeks ago, Basemark had finally released their new Basemark GPU version 1.2, which now included some bug fixes in the workloads as well as an iOS variant of the test, finally enabling cross-platform testing for mobile devices. After some internal validations, I’ve deemed it worthy to be added to our GPU suite. I’m using a custom mode at 1440p at medium settings to have it be a little more stressing in terms of the workload.

Basemark GPU 1.2 - Medium 1440p - Off-Screen / Blit

In this new test, we see relatively familiar scaling results, with things being quite on par between the Snapdragon and Exynos SoCs when it comes to their peak performance figures. It’s to be noted just how far ahead Apple’s GPUs are in this test, essentially posting figures almost 2 generations ahead.

Throttling on the Snapdragon 865 S20 Ultra is ok, only losing 22% at thermal equilibrium. The Exynos 990 S20+  was more disappointing, with performance barely better than that of the S10+ last year. The Exynos S20 Ultra again behaved very differently and for some odd reason throttled even more, actually ending up noticeable slower than last year’s model. At only 31% of peak performance, that’s some atrocious performance degradation, probably amongst the worst we’ve ever seen.

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

Moving onto GFXBench, we arrive on a familiar playing field. Peak performance of the two chips is identical, however the Exynos chip throttles significantly more. The sustained performance results here are horrible for the Exynos 990 as it’s faring worse than what we had measured on the Exynos 9820 in the S10+.

GFXBench Aztec High Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone 11 Pro (A13) Warm N7P 26.14 3.83 6.82 fps/W
iPhone 11 Pro (A13) Cold / Peak N7P 34.00 6.21 5.47 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 20.35 3.91 5.19 fps/W
iPhone XS (A12) Warm N7 19.32 3.81 5.07 fps/W
Reno3 (Dimensity 1000L) N7 11.93 2.39 4.99 fps/W
iPhone XS (A12) Cold / Peak N7 26.59 5.56 4.78 fps/W
Mate 30 Pro (Kirin 990 4G) N7 16.50 3.96 4.16 fps/W
Galaxy S20+ (Exynos 990) 7LPP 20.20 5.02 3.59 fps/W
Galaxy S10+ (Snapdragon 855) N7 16.17 4.69 3.44 fps/W
Galaxy S10+ (Exynos 9820) 8LPP 15.59 4.80 3.24 fps/W

Looking at the power measurements of Aztec high, there’s quite the big efficiency differences between the two SoCs. We had already noted that the new Qualcomm Snapdragon 865 had beat our expectations in terms of power efficiency here, sporting very big upgrades compared to the S855. The Exynos 990 on the other hand is quite disappointing in its advancements. It’s a bit better in terms of efficiency, due to it achieving higher performance, but it comes at a higher power cost.

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

GFXBench Aztec Normal Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone 11 Pro (A13) Warm N7P 73.27 4.07 18.00 fps/W
iPhone 11 Pro (A13) Cold / Peak N7P 91.62 6.08 15.06 fps/W
iPhone XS (A12) Warm N7 55.70 3.88 14.35 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 54.09 3.91 13.75 fps/W
iPhone XS (A12) Cold / Peak N7 76.00 5.59 13.59 fps/W
Reno3 (Dimensity 1000L) N7 27.84 2.12 13.13 fps/W
Mate 30 Pro (Kirin 990 4G) N7 41.68 4.01 10.39 fps/W
Galaxy S20+ (Exynos 990) 7LPP 49.41 4.87 10.14 fps/W
Galaxy S10+ (Snapdragon 855) N7 40.63 4.14 9.81 fps/W
Galaxy S10+ (Exynos 9820) 8LPP 40.18 4.62 8.69 fps/W

We’re largely seeing the same scaling in Aztec Normal, with the Snapdragon variant leading in power efficiency by 35%.

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 11 Pro (A13) Warm N7P 100.58 4.21 23.89 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 88.93 4.20 21.15 fps/W
iPhone 11 Pro (A13) Cold / Peak N7P 123.54 6.04 20.45 fps/W
iPhone XS (A12) Warm N7 76.51 3.79 20.18 fps/W
Reno3 (Dimensity 1000L) N7 55.48 2.98 18.61 fps/W
iPhone XS (A12) Cold / Peak N7 103.83 5.98 17.36 fps/W
Mate 30 Pro (Kirin 990 4G) N7 75.69 5.04 15.01 fps/W
Galaxy S20+ (Exynos 990) 7LPP 85.66 5.90 14.51 fps/W
Galaxy S10+ (Snapdragon 855) N7 70.67 4.88 14.46 fps/W
Galaxy S10+ (Exynos 9820) 8LPP 68.87 5.10 13.48 fps/W
Galaxy S9+ (Snapdragon 845) 10LPP 61.16 5.01 11.99 fps/W
Mate 20 Pro (Kirin 980) N7 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
Galaxy S8 (Exynos 8895) 10LPE 42.49 7.35 5.78 fps/W

Manhattan 3.1 also isn’t kind to the Exynos S20. The worst figure here is the fact that these S20 variants are barely any faster than the S10 in their sustained performance figures, meaning there’s zero generational improvements.

GFXBench T-Rex 2.7 Off-screen

GFXBench T-Rex Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone 11 Pro (A13) Warm N7P 289.03 4.78 60.46 fps/W
iPhone 11 Pro (A13) Cold / Peak N7P 328.90 5.93 55.46 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 205.37 3.83 53.30 fps/W
iPhone XS (A12) Warm N7 197.80 3.95 50.07 fps/W
iPhone XS (A12) Cold / Peak N7 271.86 6.10 44.56 fps/W
Galaxy 10+ (Snapdragon 855) N7 167.16 4.10 40.70 fps/W
Reno3 (Dimensity 1000L) N7 139.30 3.57 39.01 fps/W
Galaxy S20+ (Exynos 990) 7LPP 199.61 5.63 35.45 fps/W
Mate 30 Pro  (Kirin 990 4G) N7 152.27 4.34 35.08 fps/W
Galaxy S9+ (Snapdragon 845) 10LPP 150.40 4.42 34.00 fps/W
Galaxy 10+ (Exynos 9820) 8LPP 166.00 4.96 33.40fps/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
Mate 20 Pro (Kirin 980) N7 135.75 4.64 29.25 fps/W
Galaxy S8 (Exynos 8895) 10LPE 121.00 5.86 20.65 fps/W

Finally, in T-Rex, things are again quite horrible for the Exynos chip. Sustained performance is a little over half the peak performance figures as the chip suffers from major thermal throttling. Looking at the power draw, we’re reaching an awful 5.63W which is notably worse than the Exynos 9820.

Meager 3D Upgrades – Horrible Exynos Experience

Neither the Snapdragon 865 nor the Exynos 990 variants of the S20 are particularly impressive when it comes to GPU performance.

Starting off with the Snapdragon 865, Qualcomm did excellent in terms of their power efficiency and managing to reduce total power consumption compared to the Snapdragon 855. However the chip is still being curb-stomped by last two generation of Apple SoCs, and there's a lot of catching up to do in this regard.

From a device-standpoint, the Snapdragon S865 S20 barely performed any better than some of the more gaming optimized Snapdragon 855 devices from last year. The silver lining here is that both variants of the phones have outstandingly good thermal characteristics, and are usually not allowed to exceed around 42°C peak skin temperatures.

The Exynos 990 S20 variants are an outright disaster in their gaming performance. The best-case scenario here is that the new phones barely match last year’s Exynos 9820 in sustained performance, with the S20 Ultra behaving extra weirdly and sometimes falling even further behind in performance than that.

For attentive readers who noted the MediaTek Dimensity 1000L in the tables, that’s because I wanted to give some sort of notion of the Mali-G77 in a different SoC. That unfortunately didn’t help too much, as the performance points of the two chips are far too apart to come to any conclusion. What’s clear here is that SLSI clocked the GPU very high to match the peak performance figures of the Snapdragon 865, but it comes at the great cost of higher power consumption at those high frequencies.

The results of the Exynos 990 here reminded me of those of the Kirin 960 and Kirin 970 a few years back. Those parts also came out with some inexplicably horrible power figures, which I've since then heard that the matter was blamed on the use of beta GPU RTL as well as early process PDKs. Ultimately, whether it's due to Samsung's 7LPP process node or the implementation of the Mali-G77 GPU IP, the end result is that the Exynos 990 here just stinks, and those variants of the S20 have to make due with a second-tier experience.

Machine Learning Inference Performance Display Measurement
Comments Locked

137 Comments

View All Comments

  • MAGAover9000 - Tuesday, April 7, 2020 - link

    I have the s10+. Fantastic device. Very happy
  • id4andrei - Friday, April 3, 2020 - link

    No need for me to praise this review any longer. Still, I must nitpick. The 3dmark GPU test always has caveats in your reviews. Drop it if you feel it is detected by OEMs or it's a false GPU test like the physics one.

    On web tests. I read on wiki that JetStream is an Apple made test, literally. Wouldn't you say that's a big caveat when testing against ios? Similarly Speedometer is developed by the webkit team at Apple. With Android webview based on Blink, not webkit, wouldn't Android smartphones be at a disadvantage against iphones? I don't see Kraken(Firefox) or Octane(Google) being used.

    Kraken would actually be neutral to both. Other 3rd party tests might be Testdrive(Microsoft) or Basemark.
  • Andrei Frumusanu - Friday, April 3, 2020 - link

    I don't think that the fact that the WebKit team made those tests is a valid argument against using them. You can go and read the source JS yourself if you wish, and they're industry accepted benchmarks. Both Kraken and Octane are ancient and outdated and we dropped them just like we dropped SunSpider of the early days.
  • id4andrei - Friday, April 3, 2020 - link

    Thank you for the prompt answer.
  • s.yu - Friday, April 3, 2020 - link

    Thank you Andrei, again the most comprehensive and reliable set of samples anywhere!
    There seems to be considerable sample variation again (last time with Samsung was the main module since S9 with the variable aperture) in the UWA, S20+E and S20UE should have absolutely identical UWA performance but the S20UE seems to have far worse sagittal resolution than the S20+E, and Samsung's processing isn't that good in the first place, considering the 12MP 1.4μm could produce incredibly sharp pictures as that been the specs of the Pixels' main module for generations.
    I don't regret their switch to f/1.8 because the old module that went up to f/1.5 wasn't sharp wide open, especially in the corners, but a further two stops' variation to f/3.3 could be useful for more DoF in closeups provided inserting that physical aperture into the tiny module doesn't compromise the optical design otherwise.
    This time around the E seems to generally outperform the S, except in color as E doesn't seem to have proper color fidelity...almost as if chroma NR is set too high even in broad daylight, and the "hybridization" of the digital zoom, in which the E clearly uses a smaller portion from the periscope's readout than the S in the resulting merge. Speaking of the zoom, S20+ still performs slightly worse at 2x(16MP readout) than S10's native 12MP, though the difference is small and could be down to lens variation. Considering S10U's Z height, they could've easily fixed the S20U like Xiaomi, going 1/2.3" f/2 12MP with the 2x. Xiaomi used it despite a 4-1 bin, all the more reason to use it with a 9-1 bin. S20U's corner performance at 3x would also be much improved.
    Regarding the comparison with the Fuji though, I suspect your unit has trouble focusing to infinity correctly, because the train and forest samples show clear superiority of the Fuji's zoom. I especially recognize that kind of slight haziness as being very responsive to dehaze and low radius sharpening in LR and would result in far more detail with extraction in post. Also, with an ILC, there's always stopping down a little for more sharpness and more DoF.
    Regarding the full res modes, it's not worth storing 108MP of data with the CFA asking for a 9-1 bin, of course the 64MP would be better, without the RAW it's hard to say for sure, but the 64MP seems to be quad bayer.
  • s.yu - Friday, April 3, 2020 - link

    I don't agree with your remark about the night comparison with Mate30P though, the UWA is not "UW" so it has better image quality, that's true, and the night mode of the Mate30P is far superior, that's also true, but not auto mode, nor any aspect of the telephoto as it's clearly using a crop of the main for 3x. Samsung does attempt to use the 4x for telephoto and although there's a significant issue of chroma noise, it's far sharper than Mate30P's crop, with at least twice 3 times the effective resolution in night mode. With S20U you could also crop out a single shot 3-4x of similar brightness to the Mate30P crop, but it's just a crop.
    As for the potential of P40P surpassing S20U, that model operates on a 9.4MP crop by default, interpolated to 12.5MP which clearly has consequences. In daylight it's often a regression compared to P30P (much less match Mate30P), and in night shots using the current firmware it has severe color issues of rendering large portions of the scene as a crimson red, so it's hard to say at this point too.
  • s.yu - Friday, April 3, 2020 - link

    Oh, there's exception of the Mate30P auto mode in the last sample, but the night mode isn't constantly superior either.
  • RealBeast - Friday, April 3, 2020 - link

    I've been looking forward to getting one of these, not sure which yet. The fly in the ointment now is that I won't see my Mom (who gets my old S9+) until the Fall due to the whole COVID problem, not to mention less income. That will weigh heavily on sales of what is otherwise an amazing looking phone for me.
  • 29a - Friday, April 3, 2020 - link

    How large are the picture file sizes created by this thing?
  • BedfordTim - Sunday, April 5, 2020 - link

    The same size as any other 12MP camera. They will depend on content, hdr, motion and compression options but I would expect about 36MB for a raw image and 8MB for a high quality jpeg.

Log in

Don't have an account? Sign up now