Video Performance

While still images probably constitute a significant portion of what a smartphone camera is used for, video performance still remains important as anything moving or with a notable sound requires video or something that is but quite isn’t a video. In order to test video performance, we use direct comparison with two devices recording simultaneously on the same rig.

iPhone 7 Video Encode Settings
  Video Audio
1080p30 16 Mbps H.264 High Profile 86 Kbps, 44.1 KHz AAC
1080p60 25 Mbps H.264 High Profile 86 Kbps, 44.1 KHz AAC
4kP30 48 Mbps H.264 High Profile 86 Kbps, 44.1 KHz AAC
1080p120 38 Mbps H.264 High Profile 86 Kbps, 44.1 KHz AAC
720p240 38 Mbps H.264 High Profile 86 Kbps, 44.1 KHz AAC

Before we get into the actual results I want to discuss the video encode settings. It's kind of interesting to see how the iPhone only records mono audio, which I suspect is a function of not having enough microphones to do noise cancellation and useful stereo recording. It's also interesting to see how Apple can actually encode AVC High Profile for all video – including 4kP30 at 48 Mbps – which is more than I can say for a number of high-end flagships this year. This suggests that the encode blocks are capable of keeping up without any strange problems.

1080p30 Video

Looking at 1080p30 video I'm just profoundly disappointed by how high-end Android devices perform in comparison. The state of affairs here is so depressing there's really no reason to compare 1080p60, 4K, or slow motion capture because it's clear to me that something is just fundamentally broken (or consistently misconfigured) with Snapdragon 820's encode blocks. Even casual examination reveals massive macroblocking any time the sky comes into view, which is something we've consistently seen with the HTC 10, Galaxy S7, LG G5, and OnePlus 3. Other than this, the LG G5 and Galaxy S7 both have extremely oversaturated color rendition which just doesn't represent reality. The iPhone 7 and 7 Plus clearly have better output than any other Android device. The iPhone 7 Plus and iPhone 7 with the 28mm focal length camera perform quite similarly to the iPhone 6s Plus other than some improvements in dynamic range and noise reduction, so I'd refer back to the iPhone 6s review for those interested in learning about how the iPhone 7 stacks up because the state of the art in Android has not progressed since then.

OIS Video

In the interest of trying for completeness anyways, I also tested image stabilization performance. Relative to the Galaxy S7 we continue to see how the lack of software stabilization to go with the hardware stabilization leads to really shaky and jerky footage when recording while walking. The iPhone 7 still shows large motions, but it's much smoother and also handles wind noise better. The HTC 10 is much more competitive with the iPhone 7 here but I would say that Apple's software stabilization appears to be slightly better and the lack of PDAF on the HTC 10 camera is definitely noticeable in the focus transition testing but it's important to keep in mind that the HTC 10 has no software stabilization if you enable 4K video. The LG G5 has basically all of the same traits as the Galaxy S7 in this test and needs improvement in all of the same areas as a result. If you're upgrading from either the iPhone 6, 6 Plus, or 6s you're going to also see a major improvement as the iPhone 7 and 7 Plus both have OIS which works with the software stabilization to maximize the reduction in hand shake when capturing video.

I also went ahead and tested the iPhone 7 Plus with some quick footage to see what difference it provides when capturing video, and it's fairly obvious that the secondary camera lacks OIS but also dramatically increases captured detail which makes it useful for static shots where you can avoid inducing hand shake but its utility rapidly decreases in low light or high hand shake conditions.

Overall, the iPhone 7 and 7 Plus remain the best phones on the market for video capture. Strange issues with Snapdragon 820 video encode blocks mean that all Snapdragon 820 devices are just barely passable for video capture. Considering how OEMs have had almost a decade to get this right, it is truly incredible that phones costing 600 US dollars still have these obvious problems, and that Apple remains among the few to get it right.

Still Image Performance Software UX: iOS 10 and Haptics
Comments Locked

377 Comments

View All Comments

  • GTRagnarok - Monday, October 10, 2016 - link

    Yeah, that statement made no sense. He said it was fairly thick relative to the iPhone 6S which was thicker at 7.1mm so...???
  • lilmoe - Monday, October 10, 2016 - link

    The whole thing about thickness and "freezing motion" made no sense. He probably didn't have anything else to prove his "point" that iPhones faced little competition in 2014/15.

    The context of AT reviews has been broken for a while. Just like someone else here suggested; they should only focus on reviewing parts of tech in which they have more knowledge about and accurate tools to measure. Hopefully using better "tools" than browser benchmarks (!).
  • Ryan Smith - Tuesday, October 11, 2016 - link

    For better or worse, you will be seeing browser benchmarks for years to come. As few developers care to tackle cross-platform system benchmarks, we aren't left with very many tools to use. Never mind the fact that browser performance is still a major part of the overall smartphone experience.
  • lilmoe - Tuesday, October 11, 2016 - link

    Your perception and truth and credibility is up to you. But if it is worth anything, I'd suggest you change the title/section to "Mobile Safari Performance (vs Mobile Chrome)", or scratch it off entirely.

    What you're currently doing is provide false information (and lies) to your readers, unless the purpose of this site has changed to "providing fanboys and fools with fuel for their petty flamewars". You're making yourself look like an Apple shill, or worse, you're providing a cover up for Google's failings in optimizing their browser and platform for the latest Cortex and custom cores.

    If you have no tools for accurate measurement, like you just stated, then you should skip this section all together or narrow it down to Apple SoC comparisons ONLY, with a big fat disclaimer explaining why that data CANNOT be applied/compared with other SoCs and platforms.
  • lilmoe - Tuesday, October 11, 2016 - link

    ***Your perception *of* truth and credibility is up to you
  • blackcrayon - Tuesday, October 11, 2016 - link

    Yes, they should have included GeekBench as a cross platform SoC test. THAT would have shown that... oops... The A10 still blows away everything else in single core by a huge margin.

    From what I've seen, using other browsers on Android doesn't help much. But if you'd like to point to some benchmarks we're all eyes.
  • UtilityMax - Wednesday, October 12, 2016 - link

    The smartphone SoCs have reached a point where their performance differences are pretty much superfluous. Yeah, we all know that the Apple SoC is significantly better than Snapdragon 820. But that's just like saying that if you bring to a race track Lamborghini, Ferrari, and McLaren and race them, then you will find out that one car is fast than the rest, which is somewhat irrelevant outside of such racetrack. The truth is that smartphone SoCs have gotten to be fast enough that a three year old Nexus 5 is still fast and more than adequate for day to day duties. On the other hand, apple's deletion of the headphone 3.5mm jack and its refusal to allow the users to upload the multimedia into the device bypassing the iTunes GARBAGE is a big loss. For these reasons, in my book, a 200USD Moto G is still a lot better than an iPhone 7, even if the iPhone 7 was also sold for 200USD.
  • ex2bot - Wednesday, October 12, 2016 - link

    I can get the multimedia to my Apple phone without the iTunes GARBAGE. Maybe you have a different model from me or something. ;)
  • CloudWiz - Thursday, October 13, 2016 - link

    Which is why both XDA and Anandtech have found that the Note 7 was incredibly slow to load apps in Discomark, which definitely impacts daily usage.
  • ex2bot - Wednesday, October 12, 2016 - link

    Forget GeekBench. Stick with Antutu. More representative.

Log in

Don't have an account? Sign up now