GPU Performance

Device gaming performance, or better said, SoC GPU performance, is dictated by a trifecta of characteristics: First of all, the actual GPU microarchitecture and how performant and efficient it is, in this case the Adreno 640 of the Snapdragon 855 should fare very well. Secondly, the device vendor needs to employ a good physical design of the phone, able to dissipate heat away from the SoC to the body of the phone, enabling the GPU to operate at higher frequencies without throttling. Lastly, it’s a question of how the vendor actually tunes the software and the throttling levels of the phone, choosing target limit skin or silicon temperatures.

As a prelude to the test results, I’ll outright say that the Oppo Reno 10x behaves near identical to the OnePlus 7 Pro which we tested earlier this summer, meaning it has nearly no thermal throttling at all on the GPU.

Whilst the OnePlus 7 Pro was able to maintain peak performance metrics for long durations, at a cost of quite high device body temperatures, the Oppo’s long-term performance came with the gigantic caveat that it hit device thermal limits and shut down the tests.

I’ve become increasingly wary of this behaviour as I’ve encountered it in the past with several devices, Huawei in the past ended up in this situation by disabling their thermal throttling mechanisms when detecting benchmark applications, essentially cheating the behaviour and score. I’ve also encountered it early on last year on the Galaxy S9+, although Samsung claimed this was a bug and the behaviour disappeared with subsequent firmware updates.

The problem for me is that I can’t accurately determine if this is just a lack of oversight and software optimisations on the part of Oppo, or rather something that’s more malicious. I’m using benchmark suites with altered application IDs in order to avoid the more common benchmark detection by vendors, but that’s not to say they haven’t found more sophisticated ways to detect the tests.

Whatever the reason might be, the scores for the Reno 10x come with a big asterisk – being that they might not accurately depict the real performance of the device, and if they indeed are valid, then the device might simply shut down the app on you at some point if you overstress it too much or are in a very hot environment.

3DMark Sling Shot 3.1 Extreme Unlimited - Physics

The one test where we don’t see any issues is in the 3DMark Physics test, here the phone and CPU do actually throttle correctly. The Reno 10x’s result isn’t too fantastic here in terms of performance compared to other devices, but we have to keep in mind that the device allowed itself to reach much higher temperatures than other phones, and that might be the reason as to why the CPU is throttling more.

3DMark Sling Shot 3.1 Extreme Unlimited - GraphicsGFXBench Aztec Ruins - High - Vulkan/Metal - Off-screen GFXBench Aztec Ruins - Normal - Vulkan/Metal - Off-screen GFXBench Manhattan 3.1 Off-screen GFXBench T-Rex 2.7 Off-screen

In the GPU-centric workloads, we see that things are quite straightforward for the Reno 10x. It just doesn’t throttle, and the phone is able to maintain the best peak and sustained performances of any other Snapdragon 855 device. Again, this come at a cost of very high device temperatures and the risk that it’ll simply overheat and close your app if you do end up overstressing it in a real scenario.

Machine Learning Inference Performance Display Measurement
Comments Locked

40 Comments

View All Comments

  • Death666Angel - Thursday, September 19, 2019 - link

    So, PCMark score better than anything else tested here? Maybe they apply some sort of cheating fix that identifies certain apks and allows more relaxed boosting behavior, which Andrei disables.
  • ElFenix - Wednesday, September 18, 2019 - link

    an honest to goodness telephoto camera in a phone, not just a marketing BS one!
  • melgross - Wednesday, September 18, 2019 - link

    No, mostly a marketing BS one.
  • Arbie - Wednesday, September 18, 2019 - link

    No headphone jack, no sale.

    FYI, on the low light page you're missing some words in "it looks like the phone 20-30% of the highlights" .
  • SanX - Thursday, September 19, 2019 - link

    How about comparing the phone with broader range cheaper phones like UMIDIGI etc with other processors. They cost 5-10x less than Apple, Google or Samsung but definitely are not 5-10x worse, not even mentioning this OPPO phone.
  • The_Assimilator - Thursday, September 19, 2019 - link

    But does it have gangnam style?
  • Psyside - Thursday, September 19, 2019 - link

    Samsung has improved the night mode DRAMATICALLY, and also not only the Exynos get on pair with the Snapdragon, but now the exynos even manage to be better in low light, holy shit.

    And also bye bye Pixel, Samsung is so better in low light now its only bested by P30 pro and (maybe?) the new iPhones.
  • JewellMWilliam - Sunday, September 22, 2019 - link

    max.pays12.
  • Jacob36 - Tuesday, September 24, 2019 - link

    Mobile eLogbook - an iPhone and Android App for users of the elogbook.org FHI Pan-Surgical Electronic Logbook. ... Once you have entered your elogbook.org login credentials, the app will ... This option can be activated in the settings page.
    https://loginsonline.xyz/
  • SanX - Wednesday, September 25, 2019 - link

    How about comparing this OPPO Reno 10x to OPPO Reno II with new Helio P90?

Log in

Don't have an account? Sign up now