System Performance

System performance of the P20 and P20 Pro should not be something where we’re expecting any surprises. We’ve covered the Kirin 970 in detail in a standalone article earlier this year, and because the chip doesn’t offer any on-paper improvements of last year’s Kirin 960 in the P10 – the improvements in system performance should come solely based on software upgrades such as in the upgrade from Android 7.0 to 8.1 respectively 8.0 to 8.1 when comparing against the results of the Mate 10 with the same chipset.

PCMark Work 2.0 - Web Browsing 2.0

Starting off with PCMark’s web browsing test, we see the P20’s perform excellently and achieving alongside the most recent S845 devices top scores. The web browsing test of PCMark is very sensitive to performance latency, meaning how fast the CPU and scheduler ramp up to higher performance states. The P20’s use a customised variant of EAS, however still rely on an interactive CPU frequency governor.

PCMark Work 2.0 - Video Editing

The video editing subtest seems to have reached a performance plateau in terms of reported scores in that we’re not really seeing much deviation anymore across devices nowadays – giving diminishing returns of actual performance with increased scores and why all devices are so tight together.

PCMark Work 2.0 - Writing 2.0

The writing subtest is among one of the most representative tests in regards to overall user experience of a phone. Here the P20’s fare extremely well again and top the charts. Only Xiaomi’s most recent Mi MIX 2S with very aggressive scheduler configurables is able to take a lead in front of the P20’s.

PCMark Work 2.0 - Photo Editing 2.0

The photo editing subtest also is a very performance latency sensitive test – the workloads comprise very short but still heavy bursts. Here scheduler and DVFS means everything and the P20’s again are at the top of the ranks only outpaced by S845 phones.

PCMark Work 2.0 - Data Manipulation

The data manipulation workload is more characterised by longer single-thread workloads that seem to be memory latency sensitive. The P20 here seems to show a slight regression – again this agrees with the synthetic benchmarks that we’ve run in the Kirin 970 article as it seems the memory controller on the K970 is less aggressive than on the K960.

PCMark Work 2.0 - Performance

Overall, the P20’s performance in PCMark is as expected- slightly above last year’s P10 more recent Mate 10’s. Both phones are among the fastest today and only more aggressive implementations of the Snapdragon 845 will outpace them in terms of snappiness.

Speedometer 2.0 - OS WebView WebXPRT 3 - OS WebView

In web tests, the P20s again fare relatively unsurprisingly, taking a small lead over the Mate 10 in Speedometer 2.0 while still showcasing the same regression compared to last year’s P10 in WebXPRT3.

System performance of the P20s wasn’t really something I had any kind of expectations – and they ended up exactly where I thought they would. HiSilicon’s choice of sticking with Cortex A73’s in 2018 makes both sense, but also gives them a disadvantage in some markets.

At the beginning of the year I had thought we’d see the Kirin 970 have to worry about large performance deltas to the Exynos 9810 – however that worry quickly dissipated as we saw the fiasco that was the Exynos Galaxy S9. The Snapdragon 845 however still maintains a very noticeable performance lead and that’s where the P20s don’t look that good against. This would still be fine if Huawei would price the P20s more competitively – however that’s not the case, and in particular with the price premium of the P20 Pro we are not checking off all the boxes for a flagship device coming in at that price bracket.

Unfortunately Huawei’s is in a bit of a bind here as their SoC release schedule is out of sync with the technology release schedule, and that’s why the P20s will have to make due in other areas to compensate for the “average” performance of the Kirin 970. Again, I want to make note that both devices are still extremely smooth and subjectively still very fast devices – it’s just that by now we have competitions that are a generational leap ahead and that needs to be taken into account.

Software UI - EMUI 8.1 GPU Performance
POST A COMMENT

80 Comments

View All Comments

  • chrnochime - Friday, June 15, 2018 - link

    Widespread success in western market? Europe maybe but North America? I doubt it. Reply
  • Retycint - Friday, June 15, 2018 - link

    Literally in the next paragraph, "put an end to Huawei’s ambitions in the US ... However, outside the US, Huawei products are as popular as ever"

    It would be good to remember that US is just one country out of 'Western countries", and in Europe Huawei is one of the top selling brands on par with Apple/Samsung.
    Reply
  • Lord of the Bored - Saturday, June 16, 2018 - link

    US is the only western country that matters, though. Reply
  • beginner99 - Saturday, June 16, 2018 - link

    Except that Europe has double the population from US... Reply
  • mkaibear - Saturday, June 16, 2018 - link

    >double the population from US...

    And a bigger economy. Admittedly the US GDP/capita is higher but...
    Reply
  • dustwalker13 - Sunday, June 17, 2018 - link

    only if you take the whole EU as a unit including the comparatively poor eastern european countries.

    central europe has the us beat in terms of gdp/capita and the total economic power of the pop of the whole eu is significantly higher than the us.

    in the end though, neither the us nor eu are as big as the asian market at this point. i find the 'merica! guys almost funny. it is like watching an old society lady bitch about that progressive lord next door, imagining her opinion had any weight still, while in reality both are sitting in crumbling castles and the world has moved on years ago.
    Reply
  • FunBunny2 - Monday, June 18, 2018 - link

    "only if you take the whole EU as a unit including the comparatively poor eastern european countries."

    well, as Crooked Hillary said (and it's true) her votes came from districts that produced 65% of GDP. The Evil One, of course, snuck in on the basis of the skewed Electoral College biased to the low information, low income, low education empty states. the poor will always be with you.
    Reply
  • peevee - Tuesday, June 19, 2018 - link

    This is simply not true.
    Districts do not produce GDP, working people do.
    And most poor and stupid actually voted for Clinton:
    <$50,000 53% Clinton 41% Trump
    $50,000-$100,000 46% Clinton 49% Trump.
    $100,000 & over 47% 47%
    https://ropercenter.cornell.edu/polls/us-elections...

    Keep telling yourself the lies though.
    Reply
  • Round - Tuesday, June 26, 2018 - link

    This is supposed to be about a second rate phone from china and gullible buyers, not politics, but oh well...

    The only problem with Crooked Hillary (besides her not being in a CIA black-site prison for the rest of her unnatural life) and her comment/assessment, is that many highly-educated, high intelligent people decided that her brand of sociopathy was too much for them. Many hard working, also intelligent, people also found her lies too hard to swallow.

    While we have a crass and utter buffoon in office now, the myth that anyone smart enough to see Hillary for the conniving waste she is must be "low information, low income, low education" is only believed - and spewed ad nauseam - by the malignantly ignorant, non-thinking masses of clueless progressives. N.B. I an NOT talking about democrats (an endangered species), I'm only talking about the worst ignorant identity-politicing progressives.

    Perhaps learning what Hillary's globalism really means to Americans and the GDP/per capita income, instead of relying upon the dishonest pablum coming from the nuts and sociopaths on the left would be worth some time. Just a thought.....
    Reply
  • peevee - Tuesday, June 19, 2018 - link

    "central europe has the us beat in terms of gdp/capita"

    False. Only tiny countries like Switzerland and Luxembourg (in the latter case because of out-of-country commuters - similar mistake like GDP per capita in the suburbia-dominated American cities incorrectly calculated by taking product produced by all people working in the cities, including commuters from suburbs (usually producing the most product by value), but dividing them only by the number of people living in the cities.
    Statistical slight of hand, one of the many employed by liars.
    Reply

Log in

Don't have an account? Sign up now