GPU Performance & Power

GPU performance of the new A14 is something that wasn’t very clearly presented during the launch of the new iPhone 12 series. Apple had first introduced the A14 within the new iPad series where it had promised performance increases relative to the A12, not the previous generation A13, and with a bit of math this resulted into a translated 8.3% generational increase which is rather smaller than we had expected given Apple’s recent GPU trajectory over the years. Furthermore, this was also the first release where Apple compared itself to the Android SoC competition which is something the company doesn’t usually do. All these factored into some rather low expectations for the GPU of the A14 – so let’s see how that pans out in practice:

Basemark GPU 1.2 - Medium 1440p - Off-Screen / Blit

Starting off with Basemark GPU 1.2, we’re seeing a 17% increase in peak performance relative to the iPhone 11 Pro and the A13 chip, which is a nice upgrade, but doesn’t tell the whole story. In the sustained performance figure after 30 minutes of running and when the phone reaches a thermal equilibrium, we see a 45% drop in performance. In this instance, it looks like the iPhone 12 Pro reached a lower sustained performance level than the iPhone 11 Pro which isn’t a great start, but that might change with differing workloads.

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

In the Aztec High test, the iPhone 12’s fare a bit better in their sustained performances, with the new chip showing a 21% increase in performance generationally. The peak performance figure is only 11% higher but generally this isn’t the figure that is important for gaming experiences on iPhones.

GFXBench Aztec High Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone 12 Pro (A14) 🔥 Throttled N5 28.36 3.91 7.24 fps/W
iPhone 11 Pro (A13) 🔥 Throttled N7P 26.14 3.83 6.82 fps/W
iPhone 12 Pro (A14) ❄️ Peak N5 37.40 5.57 6.64 fps/W
iPhone 11 Pro (A13) ❄️ Peak N7P 34.00 6.21 5.47 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 20.35 3.91 5.19 fps/W
Mate 40 Pro (Kirin 9000) 🔥 Throttled N5 27.37 5.39 5.07 fps/W
iPhone XS (A12) 🔥 Throttled N7 19.32 3.81 5.07 fps/W
Reno3 5G (Dimensity 1000L) N7 11.93 2.39 4.99 fps/W
iPhone XS (A12) ❄️ Peak N7 26.59 5.56 4.78 fps/W
Mate 40 Pro (Kirin 9000) ❄️ Peak N5 37.22 8.53 4.36 fps/W
ROG Phone III (Snapdragon 865+) N7P 22.34 5.35 4.17 fps/W
Mate 30 Pro (Kirin 990 4G) N7 16.50 3.96 4.16 fps/W
Galaxy S20+ (Exynos 990) 7LPP 20.20 5.02 3.59 fps/W
Galaxy S10+ (Snapdragon 855) N7 16.17 4.69 3.44 fps/W
Galaxy S10+ (Exynos 9820) 8LPP 15.59 4.80 3.24 fps/W

Looking at the power consumption of the new phones, we see again that both the peak and throttled performance figures of the new chip isn’t all that much different to the previous generation, as we’re seeing roughly 8% better performance at almost the same power envelope of around 3.9W. The peak power figure of the new chip seems to have been reduced this generation and that’s very much a welcome change, and that’s where the efficiency sees the largest delta to the A13.

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

In the normal setting configuration of the Aztec test, we’re seeing again a 11% increase in sustained performance generationally, and a similar 12% boost in peak performance. These are good improvements but still a bit less than we had expected given the A14’s new process node and new GPU.

GFXBench Aztec Normal Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone 12 Pro (A14) 🔥 Throttled N5 77.44 3.88 19.95 fps/W
iPhone 12 Pro (A14) ❄️ Peak N5 102.24 5.53 18.48 fps/W
iPhone 11 Pro (A13) 🔥 Throttled N7P 73.27 4.07 18.00 fps/W
iPhone 11 Pro (A13) ❄️ Peak N7P 91.62 6.08 15.06 fps/W
iPhone XS (A12) 🔥 Throttled N7 55.70 3.88 14.35 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 54.09 3.91 13.75 fps/W
iPhone XS (A12) ❄️Peak N7 76.00 5.59 13.59 fps/W
Reno3 5G (Dimensity 1000L) N7 27.84 2.12 13.13 fps/W
Mate 40 Pro (Kirin 9000) 🔥 Throttled N5 63.56 5.37 11.84 fps/W
ROG Phone III (Snapdragon 865+) N7P 58.77 5.34 11.00 fps/W
Mate 40 Pro (Kirin 9000) ❄️ Peak N5 82.74 7.95 10.40 fps/W
Mate 30 Pro (Kirin 990 4G) N7 41.68 4.01 10.39 fps/W
Galaxy S20+ (Exynos 990) 7LPP 49.41 4.87 10.14 fps/W
Galaxy S10+ (Snapdragon 855) N7 40.63 4.14 9.81 fps/W
Galaxy S10+ (Exynos 9820) 8LPP 40.18 4.62 8.69 fps/W

The power figures showcase a similar generational movement, with a slight performance increase at a slight power decrease. It’s good progression but again not quite fulfilling our expectations of a new process node bump.

GFXBench Manhattan 3.1 Off-screen

GFXBench Manhattan 3.1 Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone 12 Pro (A14) 🔥 Throttled N5 103.11 3.90 26.43 fps/W
iPhone 12 Pro (A14) ❄️ Peak N5 137.72 5.63 24.46 fps/W
iPhone 11 Pro (A13) 🔥 Throttled N7P 100.58 4.21 23.89 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 88.93 4.20 21.15 fps/W
iPhone 11 Pro (A13) ❄️Peak N7P 123.54 6.04 20.45 fps/W
iPhone XS (A12) 🔥 Throttled N7 76.51 3.79 20.18 fps/W
Reno3 5G (Dimensity 1000L) N7 55.48 2.98 18.61 fps/W
Mate 40 Pro (Kirin 9000) 🔥 Throttled N5 87.31 4.98 17.54 fps/W
iPhone XS (A12) ❄️Peak N7 103.83 5.98 17.36 fps/W
ROG Phone III (Snapdragon 865+) N7P 93.58 5.56 16.82 fps/W
Mate 40 Pro (Kirin 9000) ❄️Peak N5 124.69 8.28 15.05 fps/W
Mate 30 Pro (Kirin 990 4G) N7 75.69 5.04 15.01 fps/W
Galaxy S20+ (Exynos 990) 7LPP 85.66 5.90 14.51 fps/W
Galaxy S10+ (Snapdragon 855) N7 70.67 4.88 14.46 fps/W
Galaxy S10+ (Exynos 9820) 8LPP 68.87 5.10 13.48 fps/W
Galaxy S9+ (Snapdragon 845) 10LPP 61.16 5.01 11.99 fps/W
Mate 20 Pro (Kirin 980) N7 54.54 4.57 11.93 fps/W
Galaxy S9 (Exynos 9810) 10LPP 46.04 4.08 11.28 fps/W
Galaxy S8 (Snapdragon 835) 10LPE 38.90 3.79 10.26 fps/W
Galaxy S8 (Exynos 8895) 10LPE 42.49 7.35 5.78 fps/W

Depending on the workload, the generational performance increases can be even smaller, as here in Manhattan the performance increase in a throttled state is only 3% better for the new A14 based iPhone, with also a minor power decrease at this state.

GFXBench T-Rex 2.7 Off-screen

GFXBench T-Rex Offscreen Power Efficiency
(System Active Power)
  Mfc. Process FPS Avg. Power
(W)
Perf/W
Efficiency
iPhone 12 Pro (A14) 🔥 Throttled N5 260.28 4.08 63.97 fps/W
iPhone 11 Pro (A13) 🔥 Throttled N7P 289.03 4.78 60.46 fps/W
iPhone 12 Pro (A14) ❄️ Peak N5 328.50 5.55 59.18 fps/W
iPhone 11 Pro (A13) ❄️ Peak N7P 328.90 5.93 55.46 fps/W
Galaxy S20 Ultra (Snapdragon 865) N7P 205.37 3.83 53.30 fps/W
Mate 40 Pro (Kirin 9000) 🔥 Throttled N5 147.13 2.92 50.38 fps/W
iPhone XS (A12) 🔥 Throttled N7 197.80 3.95 50.07 fps/W
ROG Phone III (Snapdragon 865+) N7P 224.48 4.92 45.60 fps/W
iPhone XS (A12) ❄️Peak N7 271.86 6.10 44.56 fps/W
Galaxy 10+ (Snapdragon 855) N7 167.16 4.10 40.70 fps/W
Reno3 5G (Dimensity 1000L) N7 139.30 3.57 39.01 fps/W
Mate 40 Pro (Kirin 9000) ❄️ Peak N5 235.04 6.11 38.46 fps/W
Galaxy S20+ (Exynos 990) 7LPP 199.61 5.63 35.45 fps/W
Mate 30 Pro  (Kirin 990 4G) N7 152.27 4.34 35.08 fps/W
Galaxy S9+ (Snapdragon 845) 10LPP 150.40 4.42 34.00 fps/W
Galaxy 10+ (Exynos 9820) 8LPP 166.00 4.96 33.40fps/W
Galaxy S9 (Exynos 9810) 10LPP 141.91 4.34 32.67 fps/W
Galaxy S8 (Snapdragon 835) 10LPE 108.20 3.45 31.31 fps/W
Mate 20 Pro (Kirin 980) N7 135.75 4.64 29.25 fps/W
Galaxy S8 (Exynos 8895) 10LPE 121.00 5.86 20.65 fps/W

Finally, T-Rex showcases no improvements on the part of peak performance figures, although it does lower power consumption, and sustained performance for some reason is lower on the newer generation iPhone, although again it showcases quite lower power consumption so it’s possible the new chip is mainly running on the efficiency CPU cores in this workload.

Reasonable Upgrades

Generally speaking, our concerns over Apple’s lacklustre marketing on the GPU side of things seem to have been warranted as the new A14 and the 5nm process node doesn’t seem to bring substantial gains this generation. Performance is a little higher, and efficiency has also gone up as well, but it’s nowhere near the levels of improvements that Apple had been able to achieve with the A12 and A13. On one side that’s pretty understandable as those two generations had made huge leaps, and on the other hand it was maybe unreasonable to expect Apple to continue to make such gigantic strides on every generation.

Overall, the new iPhone 12 devices and the A14 still offer the very best gaming performance of any smartphone out in the market, showcasing significantly better experiences than any other Android competitor, but it’s also not a major noticeable upgrade over the iPhone 11 series devices.

System Performance Display Measurement
Comments Locked

101 Comments

View All Comments

  • raemike - Monday, November 30, 2020 - link

    Apple markets this as the strongest iPhone ever, this is pure BS. The back glass on my iphone12 broke in 3 days. I was putting on boots and the phone fell about 18 inches to a concrete floor. Unbelievable. Apple wants me to pay $100 to fix!!!! The fact is that the iPhone 12, from my experience, is by FAR the most fragile iPhone I've ever owned. Very unhappy that Apple won't even fix their defect. Phone didn't last 3 full days!
  • Holliday75 - Monday, November 30, 2020 - link

    You dropped it on concrete ad you blame Apple? LOL.

    I shot my dog....its the guns fault. I ran a little old lady over. It's the car's fault.
  • AshlayW - Monday, November 30, 2020 - link

    Why did you shoot your dog? Also, you're probably going to prison now.
  • PickUrPoison - Saturday, December 12, 2020 - link

    Don’t you care about the little old lady?!?
  • cha0z_ - Tuesday, December 1, 2020 - link

    They strictly clarify it's for the FRONT glass, most likely they saved upgrading the back glass for the next year model + glass is glass, no matter how strong it is, the wrong angle/force and you are done.
  • 29a - Tuesday, December 1, 2020 - link

    What is unbelievable is you didn't have any protection on your phone and now you're bitching about breaking it.
  • techconc - Monday, December 7, 2020 - link

    "The back glass on my iphone12 broke in 3 days."

    If anything, that shows you how strong the ceramic shield is, since only the front of the phone has the ceramic shield.
  • zanon - Monday, November 30, 2020 - link

    Thanks for the review! I upgraded from an iPhone X to the 12 Pro Max this time around, and have been kicking the tires for a week and a half or so now. SO wanted a 12 Pro regular, and a friend got the plain 12. These days I think the additive improvements over 3-5 generations is probably how must people view these things, year-over-year upgrades, or even every 2 years, seems to quite reasonably be less common (which is a good thing). Coming directly from an X, the changes are quite significant. There are obvious sticker items like improved performance and cameras. But there is also more subtle yet extreme important day-to-day use stuff like the Face ID system being a lot more responsive and accepting more angles and distances now then the generation 1 I had before. Some other disagreement, agreement, and comment:

    Disagreement: I think you really missed the mark on how big a deal the new LIDAR sensor is for pro usage (you barely gave it a sentence) and in turn as a differentiator. In fact even though it's early days and rough, I think that might be one of the most significant Pro differentiators in a long time, "pro" here as-in actual contracting to make money. I've been trying it out with magicplan and RoomScanLiDAR and already used it at a project site. Apps like magicplan previously could be paired with bluetooth laser rangers and used that way (and still can of course), but from my testing so far the new built-in lidar does an extremely close and competent job in measuring over shorter distances and details, and you get a pretty passable v1 3D point cloud too for something you get in your pocket all the time for a few hundred extra. I was able to throw together for plans and basic 3D from scratch for a historic hotel and then use that for mockups and renovation deployment way, way better than I'd ever have expected even a year ago from a handheld with this pricing (dedicated laser 3D scan units have been available for a while, and will produce better results more quickly, but are NOT cheap or pocketable). Even the simple AR Measure app suddenly no longer feels gimicky or last resort but actually is usable without my measuring tape in many cases. The whole calibration thing is gone, it's fast, and accurate in my use to at least 1/2" which for bulk is often good enough.

    Again definitely early, early days. There are obvious holes in the apps, basic things like import are non-existent, etc etc. Yet even so it's already saved me some money and time in commercial work. As far as professional usage goes, it feels like a bigger leap forward in terms of the phone itself being a tool (vs an interface to other tools) than many things before.

    Agreement: I'm glad you highlighted that the new sharp edges are a definite step back in ergonomics. It was what struck me most immediately of course in terms of out-of-box experience. Maybe it works well for the Mini, but even on the regular (let alone the Max) the hand feel stinks compared to the iPhone X/XS/11. The edges really dig in and add to fatigue.

    However, some of that can be mitigated with a case, and that leads me to the comment that I strong suspect Apple is now designing the iPhone at least in part around case use. It's been a while since I last looked it up, but in a previous discussion we did some research and it looked like at least 75% (yes, 3/4) of phone users use cases. They're a big source of personalization, not just in terms of looks but adapting the phone to various personal use cases. Obviously drop/scratch protection to a customizable degree depending on whether someone tends to have accidents or works/hobbies in heavier duty environments, but also more exotic stuff like camera lens/telescope attachment (or just plain extra battery).

    That being the case (harhar), there a reasons to design the phone with that in mind. The camera bump for example, they extra z-distance is necessitated by the camera modules. But of course Apple could just make the whole phone thicker so that it was still smooth. And if it was expected the phone would be used bare, that'd make sense. But if it's expected it'll usually be in a case, it makes *more* sense to have it as it is now, because the available mm means the end result is something that provides whatever else the owner wants but being thinner, lighter and flush on the back than if it was a case over an already thickened device.

    As someone who has never dropped their expensive handheld stuff in 30 someodd years now and previously never bothered with a case, I do kind of miss good old naked devices. But I can't argue with the numbers either. And for the first time with the 12 I feel like a case is a requirement, not just a nice-to-have, it's too sharp and too slippery without it. Of course this makes the use of steel in the Pro even dumber, extra weight for absolutely nothing. I wonder why Apple didn't use titanium instead, they used to do a lot with that material and it seems like it'd let them claim a different "pro" material without so much weight. Oh well.

    Incidentally I think the Max might be the least popular model this time around. I was lazy about ordering, didn't bother for a few days. Yet it shipped almost instantly, while I know people who ordered regular 12s/12 Pros much more quickly than me (same channel/phone company) who are still waiting. Would be interesting if Apple breaks things out, the Mini and regular are really compelling this time (which is as it should be!).

    Thanks again for the review.
  • zanon - Monday, November 30, 2020 - link

    Also to add: CreateML (https://developer.apple.com/documentation/createml... is starting to get really impressive, and that Apple has been expending effort to allow it to all stay local rather than dependent on cloud services is useful too. Given the review points out the minimal GPU improvements, it seems like there a implications worth considering given that Apple has chosen to spend a lot of silicon budget on specialized stuff, NN etc instead. At some point I'd like to see more bench marks and articles investigating how that stuff is getting used (even basic simple first party stuff like photo facial recognition), the performance, and what kinds of network dependency and privacy the various iOS/macOS/Android/Windows/Linux implementations have. It's going to be more challenging to to see where silicon budget is yielding gains, but some of that may ultimately matter more than raw CPU/GPU in a lot of day to day and application specific use.
  • name99 - Monday, November 30, 2020 - link

    It's always difficult to know just how much of Apple's timing is planned and how much is luck.

    I think it's definitely planned that the MAC's this year have the same appearance as always, to reassure most buyers that while techies might get excited about the new internals, this is the same familiar mac as always.

    BUT I think it's also clear that the CPU/SoC this year had a very different set of priorities from performance, that it was basically more of the same. Changed where that was easy to do given the new process characteristics, but the emphasis on the stuff required for the mac, secondarily on peak power. I raise this because the corollary to the point I made above is that the best time to change the appearance is precisely when the internals are undergoing their least exciting change...

    If I had to guess, my guess would be that Apple has established the line (mini, maybe refreshed every two years; mainstream; pro; pro max) and the design language, for the next four years or so. Get used to these sizes and the squared off edges!

    On the flip side the most essential Mac-relevant SoC changes are done. Next year may again be "disappointing" in that the leading edge team will have the many core high end machines as its priority, so while the A15 SoC won't exactly be phoned in, it may still be less than we might hope.
    On the other hand, as soon as the A16 Apple may be willing to say "OK, Rosetta2 is over for new machines. If a vendor hasn't ported by now, it's not worth our time to keep coddling them". Being able to drop the compatibility stuff [memory ordering, 4kB subpages] will help a bit (less complexity always helps some), and they'll finally be past the huge effort of the transition, so once again performance can get top billing.

    Another data point for my theory (this is the appearance for the next four years or so), when do we get an Apple modem? Maybe 2022? Once again that will come in a shell that looks absolutely familiar, like an iPhone has looked for years, so that the fact of getting new technology that the techies are all chattering about won't seem strange and unnerving; what you buy will feel like last year's iPhone.

Log in

Don't have an account? Sign up now