System Performance

System performance of the new iPhones should be again excellent given the hardware advancements on the part of the new A14 chip. Alongside iOS14, we should really see no problem with everyday tasks on the phone.

Unfortunately, our benchmark suite for iOS here is still relatively barren, and we have to mostly rely on web browser benchmarks – which isn’t all too much of an issue given that’s the heaviest and most demanding every-day workload for mobile devices.

Speedometer 2.0 - OS WebView

In Speedometer 2, the new A14 showcases a large performance boost of 30% for the new iPhone generation. This is likely both due to the clock frequency increase of the new CPUs as well as the fact that the new microarchitecture has in particular a larger amount of FP/SIMD resources available to itself.

Usually at this point we would point out that it’s an apples-to-oranges comparison to other devices in the chart, and that Apple’s Nitro JavaScript engine could simply be much superior to Google’s V8 engine, but the new release of the Apple Silicon Macs where we could finally compare Safari versus Chrome has shown very little performance discrepancy, meaning the performance here is actually due to the CPUs themselves.

Having that in mind, it means that Apple’s performance advantage over Android devices has grown even bigger this generation, with little hope for upcoming Cortex cores to catch up with such a gap.

JetStream 2 - OS Webview

JetStream 2 has heavier workloads and also make uses of WebAssembly, averaging out a smaller performance increase for the new A14 chip, but still showcasing class-leading performance amongst the competition.

WebXPRT 3 - OS WebView

WebXPRT is a browser workload that tries to mimic real-world interactions and workloads. The new iPhone 12s here actually show very little progress in terms of performance.

This small progress in WebXPRT is actually quite representative of my overall impression of responsiveness of the new iPhones: The A13 and iPhone 11 devices were already so performant and responsive that the actual speed limit for user interactions nowadays are just OS animations and just general SoC DVFS, the latter which Apple has already optimised to an ideal operation a few years back.

I would be lying if I were to say that I noticed that the new iPhone 12s are any faster than the iPhone 11 in everyday usage, but that’s simply because these are already are outstandingly fast devices.

The Apple A14 SoC: Firestorm & Icestorm GPU Performance & Power


View All Comments

  • name99 - Monday, November 30, 2020 - link

    It's quite possible that world LPDD5 production capacity was not yet large enough?
    My GUESS is that the memory controller can handle LPDD5, and the M1X machines will ship with it (and those will be in low enough volume that it's not an issue).

    I was surprised that the M1 Macs did not, but presumably Apple concluded it made no engineering sense -- they are clearly fast enough as is, so why not hold something in reserve to make the M1X machines look even better when they ship?
  • vFunct - Monday, November 30, 2020 - link

    Why do you do camera tests without a SINGLE shot of a human being? You know, because how everyone uses phone cameras to take pictures of people at parties and other social events?

    Are you unable to find anyone willing to have photos taken?
  • michael2k - Monday, November 30, 2020 - link

    Taking pictures of people aren't actually important for the test scenarios:
    Daylight evaluation is measuring the sensor and software in their ability to balance contrast, dynamic range, shadows, details, and color (and of course there were some people in those pictures)

    Low light evaluation is evaluating how well the software can enhance the image without sacrificing color information, creating unusual color casts, capturing details, and balancing the light and darker portions of the pictures.

    What you seem to be asking for is a review of">Portrait Mode? The 2018">Portrait Mode review was similarly brief (both less than 5 paragraphs!)
  • vFunct - Tuesday, December 1, 2020 - link

    "Portrait-mode" is the least important piece of information necessary here for a camera review.

    You test your subjects on people to understand how well it recreates skin tones and textures.

    There were issues with previous iPhones where it would over-smooth skin textures leading to a plastic doll-like effect, for example.
  • michael2k - Tuesday, December 1, 2020 - link

    Right, but that requires you perform a Portrait Mode test, which you just claimed was the least important piece of information here.

    You cannot perform a test of the camera's skill at rendering skin tone and texture without also executing the Portrait Mode code, since that is the code that was creating the over-smooth skin texture.
  • PickUrPoison - Saturday, December 12, 2020 - link

    That over smoothing was corrected after a few weeks with an update. Reply
  • ABR - Monday, November 30, 2020 - link

    The larger phones are real porkers. I have no desire to carry a half-pound of phone around, and that's before a case! Reply
  • name99 - Monday, November 30, 2020 - link

    My favorite color is blue.

    Maybe everyone wants to chime in with totally non-technical personal statements of choice?
  • The Garden Variety - Monday, November 30, 2020 - link

    When I'm taking photos of my pizza with my iPhone, I tend to prefer almost any kind of pizza. Meat pizzas, veggie pizzas, thin crust, thick crust, New York-style crust, whatever. Not picky about pizza. Don't really care for Chicago-style, as it's basically just dough and cheese, but other than that, I'm pretty easy to please. Reply
  • flyingpants265 - Tuesday, December 1, 2020 - link

    Wow, you are not too bright. Reply

Log in

Don't have an account? Sign up now