System Performance - Still a large(r) contrast

The performance difference between the Snapdragon and Exynos S9’s was among by biggest complaints about the latter variant. Here there’s a stark difference in software quality between what Qualcomm and S.LSI were able to deliver. Let’s see if the Note9 improves this in any way:

PCMark Work 2.0 - Web Browsing 2.0

The Exynos Note9 here unfortunately doesn’t really improve on the S9, and even shows a slight regression.

PCMark Work 2.0 - Video Editing

The video editing test further showcases the same behaviour, with the Snapdragon Note9 being in line with the S9+ result, while the Exynos Note9 is in line with the S9 result.

PCMark Work 2.0 - Writing 2.0

The writing test of PCMark is in my opinion one of the most important tests in our suite, as its results pretty much directly correspond to the actual perceived speed of a device in a lot of every-day scenarios. The test makes heavy usage of common Android APIs to achieve representative usage of common tasks such as text editing and PDF rendering.

The Exynos Note9 here seems again to showcase a slight performance degradation over the S9, but it’s all within margins of error.

PCMark Work 2.0 - Photo Editing 2.0

The photo editing test consists of small bursts of workloads making use of Android’s image processing APIs. This test’s key feature is that it is very sensitive to the responsiveness of the system, in other words, how fast the SoC can ramp up its performance.

PCMark Work 2.0 - Data Manipulation

Finally, the data manipulation test is the most telling one in terms of the differences that Samsung has made on the Exynos model: Here the Note9 performs significantly worse than the Exynos Galaxy S9, coming in with a 34% lower score.

The data manipulation test is characteristic in the way it works in that it has a significant portion of heavy single-threaded processing. What’s actually happening on the Exynos Note9 here is that Samsung is disallowing the SoC to boost to its single-core 2.7GHz mode as often as the S9 originally did, a regression that I also encountered with my custom kernels on the S9.

PCMark Work 2.0 - Performance

In terms of overall result, the Exynos Note9 falls in the ranks by several spots, now scoring even lower than last year’s Exynos 8895 S8, a not too fantastic showing.

Web browsing: less 2.7GHz – more actual performance?

The most evident result of the more prohibitive single-core booster is in the web browsing tests:

Speedometer 2.0 - OS WebView WebXPRT 3 - OS WebView

Both in Speedometer 2.0 and WebXPRT 3, the Exynos Note9 performs better than the S9 with its initial firmware. The result here is directly related to the decreased result of the data manipulation score in PCMark. As explained in our scheduler pieces, one of the reasons the Exynos S9 fared so badly in these tests is the core booster mechanism; boosting to 2.7GHz on a single big core while relegating all other threads to the small cores results in worse performance than simply if there were simply more big cores available, but at a lower clock speed. The latter scenario is what happens on the Note9 as why we see a 10% improvement over the S9.

The most low-effort band-aid

Overall, the actual changes in behaviour of the Exynos chipset in the Note9 represent nothing more than the most low effort changes possible. What Samsung has done here is just slightly change the booster mechanism in order make workloads more difficult to trigger the single-core 2.7GHz boost mode. For performance this is both beneficial as well as a regression, depending on workloads. What is more important is that the severe battery life impact of the 2.7GHz frequency is more significantly reduced through these changes, even though efficiency still doesn't match the Snapdragon 845 variant.

While performance has increased in the web benchmarks by around 10% - the overall result is still abysmal. Comparing the speed of the Snapdragon Note9 to the Exynos Note9 in just everyday usage, the Exynos still pretty much falls behind in every aspect. Samsung had a chance to improve things more drastically with the release of the new phone, but to me it just looks like another disappointment.

The Snapdragon Note9 is pretty much in line with other S845 devices: performance is a non-issue. While there are now more contrasting devices out there such as Huawei’s Mate 20’s – the Snapdragon Note9 is still a great device to use when it comes to its performance.

Introduction & Battery Life GPU Performance & Device Thermals
POST A COMMENT

69 Comments

View All Comments

  • cha0z_ - Wednesday, November 28, 2018 - link

    Actually for us the phone costs more. Euro is higher than the US $ + they also add atleast 100 on top of the us price. It's 1100 euro (1240$) vs 1000$. :) Reply
  • vivekvs1992 - Monday, November 26, 2018 - link

    Obvious question.. Why is note 8 not on any of the charts?? I'd be more interested in comparing note 8 rather than lg g6.. Reply
  • B3an - Monday, November 26, 2018 - link

    The Exynos 9810 is such a massive flop.

    I had to buy an import in order to get a SD845 version of my S9+. Hope Samsung fix future Exynos SoCs, because the gap in performance and battery life is just ridiculous and not acceptable. Infact Exynos 9810 isn't even acceptable for a high-end device.
    Reply
  • wintermute000 - Tuesday, November 27, 2018 - link

    Can't argue with the numbers, however, anecdotally, my exy S9+ feels plenty fast with decent battery. I haven't heard anyone who doesn't haunt XDA etc. actually complain about exy, even techies. So I can only imagine that the SD845 version is even better as a phone LOL Reply
  • cha0z_ - Wednesday, November 28, 2018 - link

    People don't get to hold the both variants to see the difference. Everyone who did and mind you - coming prepared for a difference into it - said that he is surprised how big it is and kept the SD variant. Reply
  • LiverpoolFC5903 - Tuesday, November 27, 2018 - link

    Some of these results for the Exynos look very dodgy, including one subtest where the 3 year old Snapdragon 820 scores higher. I wonder how accurate or representative these PC mark tests are, because that is all that the author has used to trash the Exynos 9810. Other tests/suites need to be run to validate these results. I am finding it hard to believe that an SOC with 'fat' cores will lag behind last years SOCs in many tests. Reply
  • GreenReaper - Wednesday, November 28, 2018 - link

    It's absolutely possible. Consider the current state of AMD graphics vs. NVIDIA - Vega doesn't really compete with the 1080, let alone the 2080. Then add in the fact that in mobile, inefficiency counts double, sinceit both drains the battery and causes thermals to spike; throttling that 'fat' CPU. Reply
  • tuxRoller - Wednesday, November 28, 2018 - link

    The author used other tests for graphics and threw in a few js benchmarks.
    I also think that others have responded about why having "fat cores" doesn't automatically mean better perf. Briefly,
    unless your data set is tiny, you have to make sure that stalls happen as infrequently and briefly as possible. This requires a very well executed cache, a smart memory controller, and software that is well aware of how best to make use of it all.
    There's more detail in the earlier articles about the issues with the 9810.
    Reply
  • s.yu - Tuesday, November 27, 2018 - link

    Thanks for another fun article!
    This would provide guidance for buying to more people if released earlier in the product cycle though :)
    Reply
  • AceMcLoud - Tuesday, November 27, 2018 - link

    Happy to see my iPhone 6s still holding up to 2018 Android "flagships". Reply

Log in

Don't have an account? Sign up now