GPU Performance

The A12 SoC in the iPhone XSs showcased some fantastic results with some extremely high scores. We expect the iPhone XR to perform just as well – the only unknown factor here is the thermal designs of the two phones.

The iPhone XR actually differs quite a lot in this regard to its XS siblings: The iPhone XS and XS Max employ a “sandwiched PCB” motherboard design, in which the SoC lies between two substrates. While I don’t have empirical data on this, I just have to assume that such a design is not helpful for actually dissipating heat away from the SoC to the body of the phone.

iPhone XS vs. iPhone XR PCBs (Images Courtesy iFixit)

The iPhone XR on the other hand employs more of a traditional single PCB design, such as found in past iPhone models. The SoC here is found inside an EM shield facing towards the screen assembly. I was always questioning this a design as well because theoretically there should be a slight air gap between it and the display backplane which could hinder heat transfer. Also because the SoC is facing towards the display, it also doesn’t make direct contact with the aluminium frame of the phone. As Apple’s SoC’s become ever more power hungry at their peak performance states, it is weird to see that Apple hasn’t been any more aggressive in their thermal dissipation solutions as say what we now commonly see from some higher-end Android counter-parts.

3DMark Sling Shot 3.1 Extreme Unlimited - Physics

Starting off with the physics subtest, which is mainly a CPU test within a (GPU) thermally constrained scenario, we see the iPhone XR perform nearly equally to the iPhone XS, showcasing some very good sustained performance near the peak levels of the SoC.

3DMark Sling Shot 3.1 Extreme Unlimited - Graphics 

The graphics subtest of 3DMark also doesn’t seem to drastically differ from the iPhone XS. This was one of the tests that caused problems for the XS variants as it would cause the phones to crash the GPU at peak performance, only able to complete the test when the phones were warmer and throttled more. I re-tested the XS and did manage to somehow complete a run at a higher peak performance state, however my iPhone XS Max still managed to crash with the same behaviour as back in October.

Moving onto the GFXBench suite, starting off with the two variants of the new Aztec sub-test, which represents a more modern 3D workloads:

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

The iPhone XR performed extremely well in these two tests, however we do note that there’s a difference in the peak performance showcased by the XR and the XS variants: It seems that Apple might be running the GPU at a slightly lower frequency here. This would match the slight difference in the peak 3DMark graphics score as well, and could be the reason why the iPhone XR didn’t crash in the same way the XS did.

GFXBench Manhattan 3.1 Off-screen GFXBench T-Rex 2.7 Off-screen

In the Manhattan 3.1 and T-Rex offscreen tests we see similar behaviour: the peak performance of the XR is slightly lower than that of the XS phones, however the sustained scores are higher.

One of the reason we use off-screen scores in our benchmark suite is that it allows for an apples-to-apples workload comparison across all devices. While in most cases this is a good and representative scenario for what you might expect from the thermals in actual games, this actually doesn’t seem to apply for the iPhone XR:

Because of the relatively low screen resolution on the part of the XR’s display – a sub-FHD 1792×828 – I’ve had a lot of trouble actually getting workloads to push the A12's GPU to its peak frequencies in on-screen scenarios. This causes an interesting dilemma for the iPhone XR: It has absolutely abundant GPU performance that won’t be used any-time soon. As game developers on iOS will be targeting and tuning their workloads to run smoothly on the most demanding devices of a generation, it means that games will most likely be setting their baseline as the higher-end iPhone XS Max, which has to push over twice the pixel resolution. The net result is that for any given 60fps graphics workload, the iPhone XR will run cooler and with a longer battery life than what you would experience on the XS or XS Max.

If gaming and gaming performance (and battery efficiency) is important for you, then the iPhone XR is definitely the phone to get. Not only does it deliver class-leading performance, but it also will be quite future proof in terms of performance thanks to the lower-resolution display.

System Performance Display Measurement
Comments Locked


View All Comments

  • GreenReaper - Tuesday, February 5, 2019 - link

    I don't think that's the case. LCD is more efficient than OLED unless you're looking at a largely-black screen. There have been reviews looking specifically at power usage and they show a significant increase for OLED screens.
  • FreckledTrout - Wednesday, February 6, 2019 - link

    Its both the resolution and the use of LCD since LCD has lower standby draw.
  • - Wednesday, February 6, 2019 - link

    don't think thats true.... OLED only light up pixels that needs to be lite up.. while LCD requires the whole backlight to be lite up... unless you really dim the LCD, OLED typically is more efficient.
  • - Wednesday, February 6, 2019 - link

    Proof OLED is more effecient than LCD ... from macworld
  • cha0z_ - Monday, February 11, 2019 - link

    Wrong, even the current gen LCDs are draining more power than a OLED current gen screen. The only exclusion would be a maxed out brightness, but most people don't use the phone in max brightness most of the time. Then again, I don't know what the power consumption of the XS OLED is (can't recall the Andrei tests, you can check them out in the review).

    The lower screen resolution is the reason for the long battery life, not only because the UI is using a lot less GPU to draw, but this goes to every app too - like youtube, that will run most videos at 720p while my note 9 will run them at 1440p if available. The workload for 720p and 1440p is a lot different and leads to different battery usage.Then we have games... for xs they will be it's native resolution in a lot of cases, but for the xr they will be 720p. And so on and so on... the resolution is playing a lot bigger role than: "oh, the UI is using so low on the CPU/GPU that the resolution doesn't matter." yeah, for the UI itself the difference would not be that great, but on the most consumed apps like videos/games/modern web pages - it will.
  • Wardrive86 - Tuesday, February 5, 2019 - link

    I always look forward to your in depth reviews and architecture deep dives Andrei. I love your sustained performance tests..something only found at Anandtech currently. I honestly think you are the best writer Anandtech or any other tech site has had period...with that out of the way, I have one question and a potential request. Spec2006 shows the A12 having substantially better efficiency than it's competitors, finishing the finite tests faster with less total energy used by virtue of higher performance, but having higher average power draw. How would this correlate to an infinite test such as a high end 3d game like Shadowgun Legends or Asphalt 9? Would the higher average power draw remain at the same performance level of say a Snapdragon 835/845/Kirin 980? Also a request if you ever have time with your busy schedule, Could we get a sequel article to your 2015 article "The Mobile CPU core count debate" with more modern games and apps, comparisons of OpenGLES vs Vulkan on CPUs, etc.. I know I don't ask for much!
  • Andrei Frumusanu - Wednesday, February 6, 2019 - link

    > How would this correlate to an infinite test such as a high end 3d game like Shadowgun Legends or Asphalt 9?

    For infinite workloads you just use the perf/W metric:

    Consequently if you're fps capped you can use Joules/frame but that's not a popular metric.

    > Could we get a sequel article to your 2015 article "The Mobile CPU core count debate" with more modern games and apps

    At some point.
  • zodiacfml - Wednesday, February 6, 2019 - link

    Performance reminds me of my AMD Vega 56 on reasonably low graphic settings at 4k/UHD resolution, it uses only 60 to 80 watts while gaming. Changing just one or two graphics settings shoots to its default power limit which is 180 watts.

    Anyways, I hope Apple release an iPod touch this year based from this SoC. It is terribly impressive which is good for gaming.
  • CHJ - Wednesday, February 6, 2019 - link

    I have been using the XR for the last few weeks. My previous phone was the Xperia XZ, and the one before that a 5s.

    I cannot stress enough how good the battery life is. I used to carry a 10000 mAh power brick around for a full day out just in case - that simply is not the case anymore. I can’t run this thing down below 20% (I do sometimes get close, meaning I probably will run down the XS on some days, especially taking into consideration battery degradation after some months). This has simplified my life quite a bit.

    As for the display, I was worried about it, but I honestly do not notice the lower resolution compared to the XZ in day-to-day usage. In fact, all things considered (color accuracy and True Tone), this has a better display imo. The lower resolution certainly helps battery life as well.

    The thicker bezels - this I do still notice, but the XR costs 330 USD less than the XS where I live, and a slightly worse display, the lack of a second lens and 3D Touch, and thicker bezels are acceptable compromises given the savings. And that amazing battery life, did I mention that? ;)

    All in all, I am very satisfied with my purchase.
  • CHJ - Wednesday, February 6, 2019 - link

    Some other observations:

    The weight is noticeable, and the width of the phone does not aid in getting a comfortable grip. My fingers cannot wrap over the edge when I need to reach the top half of the screen. If you have large hands, you’ll be better off, I think.

    The stereo speakers are also much improved and genuinely useful. I find them to be sufficient for watching YouTube, which is pretty much all I need from phone speakers.

Log in

Don't have an account? Sign up now