iPhone 16 Pro Review: Solid Upgrades And Future Apple Intelligence
Apple iPhone 16 Pro: Camera Testing and Performance Benchmarks
iPhone 16 Pro Cameras and Controls
The Action Button might be able to launch the camera, but that's unnecessary because on the other side of the iPhone 16 Pro sits a dedicated camera button. This is a two-stage trackpad of sorts. With the camera app open, half-pressing the button focuses the camera and invokes a zoom slider, and the button responds to a sliding gesture to set the zoom arbitrarily. On previous iPhones, that was a pinch gesture -- which still works here, at least for now. Pressing the camera button all the way takes a photo, or you can still press the big white on-screen button to take a snapshot. Long-pressing the button takes a short impromptu video, just like holding the circular on-screen button.So, now there are two ways to do the same things, and one of them is clearly superior to the other. Unfortunately, despite all the hubbub that Apple made about the camera button, it's half-baked in design -- or perhaps over-baked. Half-pressing the button is easy enough to get the extra controls on screen, but fully pressing the button is super stiff. Compared to the Nikon D5300 used to take the product photos in this review, it takes without exaggeration about three times as much pressure to actuate the button. And because our hands are fleshy, the phone actually moves in the hand when applying all that pressure, changing the framing of the photo ever so slightly. That just ruins the experience and kills the illusion that you're using a fancy mirrorless camera system. Apple just whiffed when creating this thing, and it's a shame, because it could have been fun.
Let's talk about the cameras themselves, of which there are four. The camera loadout is the third differentiating feature compared to the vanilla iPhones, and we'll point out the differences as we go. On the back, the iPhone 16 Pro has the traditional wide-angle, standard, and telephoto lenses. The main camera is more or less the same 48-megapixel sensor that the iPhone 15 Pro had. All of the examples below were captured with the default settings with no post-processing applied.
We'll start with the 12 MP f/1.9 aperture front-facing camera, which has not changed since the iPhone 12 Pro. There are extra features, to be sure, but those are software features -- improved fake depth of field for portrait mode and enhanced HDR are part of the iPhone 15 Pro's A17 Bionic and this model's A18 Pro's image processor, not something that has changed in hardware for four years. Fortunately, it still does a serviceable job for video chats, Instagram photos, and 4K 60 fps video.
By default the main so-called "Fusion" camera takes 12- or 24-megapixel photos by clustering the pixels into groups of two or four, and the quality is nice. Nicer still is the ability to capture 48 MP images in Apple's ProRAW format. By default, the iPhone 16 Pro captures 24 MP images. Just like the last couple generations of iPhones Pro, the Fusion camera can also take a center crop of 12 megapixels for what Apple calls a 2x optical zoom.
As far as we can tell, the Fusion name comes from the software and changes to the ISP in the A18 Pro, as the hardware hasn't changed since the iPhone 15 Pro. However, Photo Styles have been updated with the 16 Pro in that they're saved separately, so they can be modified after the fact, whereas with previous generations they were applied to the saved image data. The new Fusion image pipeline allows users more flexibility with postprocessing effects.
Video can be captured at 1080p or 4K at frame rates up to 120 fps. All combinations of resolution and frame rate support Dolby Vision as well. While all of the iPhone 16 Pro models can capture Apple ProRes video, only the 256 GB iPhones Pro can capture ProRes at 60 fps, while the 128 GB base model is limited to 30 fps. The file sizes are enormous, up to 1.7 GB per minute. External storage can support 4K ProRes at 120 fps, which seems kind of insane for a phone. There's a lot of ISP horsepower under the hood and Apple is doing all it can to flaunt it.
Incidentally, all of this hardware is identical between the iPhone 16 Pro and iPhone 16 vanilla. The only differences are that the cheaper devices are capped at 60 fps video and cannot capture ProRAW photos or ProRes video. We're not sure if the iPhone 16's A18 non-Pro was designed with less image processing capabilities or if this is just a software block. Here are some examples from the Fusion camera, where each photo's settings are listed in the caption.
The iPhone's low-light mode relies on long exposure times and some machine learning to stitch together several captured images. That hasn't really changed here. Below, the low light photo is pretty bright, but it had a three-second exposure time, and the resulting full-sized image is kind of blurry. The well-lit image had every light in the room on and exposure was almost instant, just like it was outdoors. Click through to see the difference.
What's brand new and exclusive to the iPhone 16 Pro is the 48 MP ultra-wide angle sensor. Previously this was a 12 MP snapper, which is still found on even the new midrange phones. This one captures 48 MP images by default, though it can be dialed down to 12 and 24 megapixels. Adjusting the zoom between 0.5x and less than 1.0x uses a crop of the sensor, which does a really nice job with macro shots as well.
Finally, the telephoto lens from the previous generations is still present. It hasn't changed at from last yearl, except that the 5x optical zoom that was previously exclusive to the iPhone 15 Pro Max has trickled down to the standard-sized iPhone 16 Pro this year (as well as the 16 Pro Max). This still has a 12 megapixel sensor and an f/2.8 aperture. Here are some samples.
Now let's take a look at what's missing -- or perhaps you could look at it as coming soon.
Apple iPhone 16 Pro Benchmarks
We're a little limited in the benchmarks that we can run on the iPhone 16 Pro. Not everything that we typically run (particularly PCMark Work for Android) has a counterpart on iOS. However, we can still run what cross-platform benchmarks that are available and for which we have competitive data, and that's just what we did.For the iPhone 16 Pro, all of these tests were performed on the stock 18.0 operating system, not a beta. The same is true for every other phone on these charts; they all were running the latest version of the operating system available at the time they were captured. For the iPhones, it's all iOS 18, and for the more recent Android devices, that means an up-to-date Android 14.
GeekBench 6 is the latest version of Primate Labs' synthetic benchmark. This test runs a series of burst-focused, real-world scenarios including data compression and encryption on both a single core and using all available hardware threads.
The iPhone 16 Pro only has six total cores, and only two of those are of the performance-oriented variety. However, it takes the top of our chart in both multi-threaded performance as well as single-threaded. And looking at that single-threaded score, it's easy to tell why. It's not on here because this is a phone review, but a single-threaded score of 3350 beats not only all the phones here, but also the M2 Pro MacBook Pro 14" that we reviewed not quite two years ago. The test tends to favor Apple hardware, but even when comparing Apples to Apples, the iPhone 16 Pro seems pretty darn speedy. Of course, that could change with the next generation of Snapdragon 8 SoCs.
Next up is GeekBench AI. It again uses real-world tests including image classification, text parsing, and object recognition. We chose to run this on the NPU of each device, rather than test the CPU. This is mostly due to the fact that purpose-built hardware like the iPhone's Neural Engine should be the most efficient way to crunch some AI numbers.
Next up is GeekBench AI. It again uses real-world tests including image classification, text parsing, and object recognition. We chose to run this on the NPU of each device, rather than test the CPU. This is mostly due to the fact that purpose-built hardware like the iPhone's Neural Engine should be the most efficient way to crunch some AI numbers.
We skipped reporting the single-precision result here because 32-bit floating point numbers are not typical of the kind of AI workload that we expect space-constrained phones to be running. The models are much larger, they require more RAM, and often that level of precision isn't necessary to get acceptable results. Also, with NPUs, the scores are way lower because they're just not built for this kind of work, which makes the chart hard to read.
At any rate, the iPhone 16 Pro comes out on top in both the Half-Precision (FLOAT16) and Quantized (INT8) data types. Half-precision is much faster on any iPhone on this chart than it is on any competing device. However, we think that for all the reasons we just mentioned, quantized data will be the bread and butter of any phone's NPU as long as the results turn out as expected, and with the quantized data it's a much tighter race here. Recent devices with Qualcomm Snapdragon 8 Gen2 SoCs like the Samsung Galaxy Z Fold6 score above last year's iPhone 15 Pro Max, though come up a bit shy of this year's iPhone. That's the kind of leap-frogging performance we like to see in a competitive space, although the new iPhone still scores a better than 15% margin of victory.
Pour one out for the Pixel lineup, though. Since they don't have Qualcomm SoCs, they're not able to use Qualcomm's QNN framework for AI, and are stuck using the stock version of Android NNAPI. These results are much, much slower than anything from either Qualcomm or Apple. The Google Tensor G4 has the same third-generation Tensor Processing Unit that the Pixel 8 family's Tensor G3 has (what a confusing series of naming conventions) but it's nearly an order of magnitude slower than the iPhone. For what it's worth, even Qualcomm SoCs score much, much lower using NNAPI as opposed to QNN, so it seems the Android framework may be to blame here. It could very well be that some software tweaks could incease these results quite a bit.
Next we'll turn an eye to graphics. 3DMark Wild Life is a tried and true mobile benchmark that uses the latest APIs on all supported platforms, so that means Vulkan on Android and Metal on iOS.
At any rate, the iPhone 16 Pro comes out on top in both the Half-Precision (FLOAT16) and Quantized (INT8) data types. Half-precision is much faster on any iPhone on this chart than it is on any competing device. However, we think that for all the reasons we just mentioned, quantized data will be the bread and butter of any phone's NPU as long as the results turn out as expected, and with the quantized data it's a much tighter race here. Recent devices with Qualcomm Snapdragon 8 Gen2 SoCs like the Samsung Galaxy Z Fold6 score above last year's iPhone 15 Pro Max, though come up a bit shy of this year's iPhone. That's the kind of leap-frogging performance we like to see in a competitive space, although the new iPhone still scores a better than 15% margin of victory.
Pour one out for the Pixel lineup, though. Since they don't have Qualcomm SoCs, they're not able to use Qualcomm's QNN framework for AI, and are stuck using the stock version of Android NNAPI. These results are much, much slower than anything from either Qualcomm or Apple. The Google Tensor G4 has the same third-generation Tensor Processing Unit that the Pixel 8 family's Tensor G3 has (what a confusing series of naming conventions) but it's nearly an order of magnitude slower than the iPhone. For what it's worth, even Qualcomm SoCs score much, much lower using NNAPI as opposed to QNN, so it seems the Android framework may be to blame here. It could very well be that some software tweaks could incease these results quite a bit.
Next we'll turn an eye to graphics. 3DMark Wild Life is a tried and true mobile benchmark that uses the latest APIs on all supported platforms, so that means Vulkan on Android and Metal on iOS.
Apple's string of wins comes to a screeching halt here as the crown goes to several Snapdragon 8 Gen3 phones, including a trio of Samsung phones like the Samsung Galaxy Flip6 and the ASUS ROG Phone 8. The iPhone is about 15% behind here, as it seems that despite Apple's talk of juicing up the GPU and including fancy ray tracing features didn't make up the difference.
GFXBench is another cross-platform benchmark and we took a look at two of the stock scenes there - Aztec Ruins and the T-Rex benchmark.
GFXBench is another cross-platform benchmark and we took a look at two of the stock scenes there - Aztec Ruins and the T-Rex benchmark.
This actually turns out even worse for the iPhone, as it falls further down the charts in both tests. The T-Rex frame rates are kind of insane; no phone has a 330 Hz display let alone the 500+ that the top contenders score. It's not a particularly complex test. On the other hand, Aztec Ruins is the most advanced test that GFXBench has, and the iPhone does a bit better there. Still, it's something like 30% off the pace set by the ASUS ROG Phone 8 Pro in its extreme performance X-mode.
Before we render our conclusion on the next page, we'll take a moment to discuss what's coming -- indeed, what was missing at launch -- on the iPhone 16 Pro. Let's talk about Apple Intelligence next.
