MacBook Pro (2023) with M2 Pro Review: Peerless power, productivity, and portability, done again
All that power, constrained by its form factor
In 2021, the world saw the true capability of Apple silicon when Apple announced the M1 Pro and M1 Max. We knew that Apple was going to build a chip that had more processing power than the M1 — but we didn’t know how much performance it would offer, and most importantly, we knew nothing about the chassis those chips would rest in. We had been getting mixed rumors about a 14-inch laptop from Apple that would add back ports, remove the touch bar, and have a better screen, but that was it — up until Apple’s Unleashed event in October. Almost 18 months later, we’re here again, after the release of the M2, M2 Pro, and M2 Max, the latter 2 of which were unveiled at a pseudo-event in the middle of January, defying the words of every leaker and their sources. The chips, unsurprisingly, sit in a new, refreshed MacBook Pro, this time, with promises of better battery life, better neural processing, and a fair boost in CPU and GPU performance. But of course, we had no information about the exact tangible advantages these new laptops would have over their predecessors, so someone had to put that to the test. Over the past week-and-a-half, I’ve done everything imaginable to these machines; I’ve put them through the most demanding synthetic benchmarks, put them through video encode sessions and Xcode compiles, and rigorous battery testing, and I’ve got answers — concrete answers about how much better these chips actually are, how they scale across versions, and how they compare with other computers Apple sells in 2023. This isn’t a normal review — it’s a review about the unique hardware capabilities of the chip inside of the untouched chassis from last year — there’s no commentary about the keyboard, or notch, or screen, or ports selection, and if you want that information, my review from last year has all of it covered. As I said, the design, materials, and external capabilities of this machine are untouched and are a mirror of the design many industry professionals have praised over the last 15 months. It’s time that we took a deeper dive into this machine — one that runs circles around the top-of-the-line Mac Studio with the M1 Ultra in some circumstances but simultaneously falls flat when compared to its precursor. It’s peculiar, it’s unique, and also largely the same; unchanged. I’ve got the base-model M2 Pro machine, I’ve done all the math, I’ve compiled all the charts, and I’ve got all the numbers in tow. This is the only review of the M2 Pro MacBook Pro (2023) you’ll ever need.
As I said before, one of the main promises that came along with the new MacBook Pros, and the M2 chips that come in them, was that there would be a substantial increase in battery life. In 2022, after the introduction of the M2 chip at WWDC and the subsequent M2 MacBook Pro, Apple made the bold claim stating that the 13-inch M2 MacBook Pro had the “best battery life of any Mac notebook.” Many reviewers have put this claim to the test, and it turned out to be true — the large battery and low-resolution screen coupled with the base M2 provided for a superior battery experience, one that was superior to even the 16-inch M1 Pro MacBook Pro — the previous champion. Now, the M2 Pro 16-inch takes its crown back, with Apple stating that it now has the “best battery life of any Mac laptop.” However, Apple’s numbers they use to back up their “claims” are usually bogus and irrelevant — Apple measures battery life for its spec sheets and comparisons in 2 ways, for the most part: they measure “video playback” performance, and “wireless web” performance. The video playback test usually entails looping a 4K video with auto-brightness on in QuickTime or the TV app, with the goal being to show how the video decode engines are extremely power-efficient (which they could be correct in asserting). The latter comparison entails opening a variety of websites (obviously, we have no idea what these websites are, how many there are, and what they’re doing on them) in Safari and measuring how long the battery lasts, which I, of course, think is a much more useful test. Still though, both comparisons are pretty useless unless you want a ballpark estimate as to how much worse (percentage-wise) the old versions are compared to the new ones. Apple claims, through these “tests,” that the machines last one hour longer compared to their predecessors, which roughs out to about a 7% improvement year-over-year. Not too shabby for a sterilized test of which we have no details about.
In my personal testing, however, I noticed a significant difference in how long the computer lasted — a difference much more substantial than what Apple claimed, to my surprise. Last year, I said the following:
I was lucky enough to get 10–12 hours of SoT on MacBook Pro doing normal things like email or word processing. However, when it came to video editing, the computer died in nearly 6 hours. Yikes. That’s not to say that this computer has inherently bad battery life. It’s just obviously not as good as the efficiency-oriented MacBook Air and consumer MacBook Pro.
Realistically, I was getting roughly about 8 hours of screen on time between charges on my MacBook Pro from 2021, juggling Safari, email, and some other (native) apps utilizing the efficiency cores — nothing crazy to see here. The same isn’t the case with the new generation: the slightly larger battery and more efficient process node of the M2 chip got me about 9–10 hours of web browsing and other productivity work, and that difference is nothing to scoff at. Most of the time, regardless of how “pro” you are, efficient, lightweight tasks are the most common thing most people do on a laptop, so an extra 2 hours of juice isn’t a bad upgrade. I was noticing that the machine let me do more work throughout the day, lasted for more days on a single charge, and was able to get me through long work sessions with just a quick 15-minute juice-up (and I don’t even have the fast charging brick). Another way I tested this was by setting a stopwatch when the laptop alerted me that the battery was running low, and I found that the new machine was able to last over 1.5 hours doing productivity work in Safari and typing up documents. Let me reiterate: the computer took 1.5 hours to drop just 10%. That’s extremely impressive, especially judging by the fact that the 2021 machine took 40 minutes or so when I did the same test on it when it was new. I’m confident in saying that both machines, regardless of configuration, provide significant, tangible, and appreciative upgrades in battery performance for lightweight tasks.
In addition to my day-to-day testing, I also observed that the laptop lasted longer when I was rendering out a video or performing synthetic benchmarks, even when the fan was cranked up above 2000 rpm. Rendering and exporting a long video 4 or 5 times would kill the old MacBook Pro instantly, but this one lasted considerably, and noticeably longer. In fact, I did all of my synthetic benchmarks (which you’ll hear about later) on battery power to test the new battery, and it got through everything but the Final Cut Pro exports at the end without dying. It’s truly exceptional battery performance with the 14-inch variant, and I can’t imagine what it’ll be like with the 16-inch.
Before I give it my full seal of approval though, I do have to reiterate that I’m using the M2 Pro variant of the MacBook Pro, not the M2 Max. If these machines are anything like last year’s, the M2 Max MacBook Pro will be slightly less efficient and will last a couple minutes less, especially if you have the 14-inch model. But in my opinion, these differences won’t really matter in the real world, mostly because of the large gains year-over-year, and the astronomical gains coming from Intel. These laptops; they last a long time. It’s pretty hard to kill them unless you’re really punishing them as Jason Snell did in his review, where he pegged the GPU and CPU at the same time, making the 16-inch run dry in 3 hours. I don’t think that’s something many of us will do most of the time.
In addition to class-leading battery life, the M2 Pro’s predecessors, and Apple silicon chips in general, are known for their excellent thermal performance. Even the monstrous M1 Ultra with 64 GPU cores runs cool when compared to previous Intel Macs. However, this is where the M2 Pro starts to… fall apart. If you were paying attention to the Apple world in August 2022, you must at least be familiar with the M2 MacBook Pro’s poor thermal performance found found by reviewers. Last year, many found that the M2 MacBook Air thermal throttled during even simple workloads, regardless of configuration and ambient temperature. However, the MacBook Air was let off relatively easily by the media and YouTube comments section due to its lack of a fan and lack of “Pro” moniker, which would say that the machine wasn’t meant for such thermally-exhaustive tasks. When I started testing the M2 Pro, I was concerned that this otherwise phenomenal machine would suffer from a chip that overheated. And well… I’m sad to say that this is one of the M2 Pro MacBook Pro’s major flaws.
While I’ll save the actual results my benchmarks for the performance section, I will discuss my thermal findings: the M2 Pro MacBook Pro thermal throttles when put under extreme stress unlike the M1 Pro one did, and the throttling was enough to cause a significant drop in performance. The M2 Pro model, which I’ll now refer to as the M2 MacBook Pro for simplicity, warmed up by a fair margin when running a 30-minute multi-core CPU render in Cinebench. By how much? The machine went up to 220º Fahrenheit, or approximately 104º Celsius on average throughout the test, and hit a maximum temperature of 232º Fahrenheit, or 111º Celsius at its peak towards the end of the benchmark. Usually, this wouldn’t be much of a big deal, especially knowing that Intel laptops frequently reach 150º Celsius when under peak workload. But Apple doesn’t ramp the fans up when the machine is heating up, even at the cost of possibly thermal throttling. And even when they inevitably do ramp the fans up when the machine screams for air, they’re not as powerful as they need to be to cool a processor that runs so hot. This wasn’t such a problem with the M1 series of laptops, but for some reason it is for the M2s. The result is a machine that warms up a lot and is much louder than its predecessor, which was infamous for being quiet and cool. To test this even further, I put both machines side by side in an air-conditioned room while performing the tests, and I could audibly hear the M2 MacBook Pro’s fans from about 10 feet away, while the M1’s was dead silent. And even after the test, the M2 laptop took considerably longer to cool off and reach room temperature when compared to the previous generation. This isn’t a good thing; it’s exactly what I worried about, and is one of my biggest problems with this machine. The M2 Pro is constrained by the chassis it’s in — in fact, I have a theory that may support this but that I, unfortunately, can’t test: if we took an identically-spec’d M2 Pro Mac mini and pitted it against a MacBook Pro, the Mac mini would win over the MacBook Pro by a long shot since the mini wouldn’t throttle. Again, this is just a theory, but the thermal chassis in the Mac mini is superior in size and power to the one in the MacBook Pro, so I wouldn’t be all too surprised.
The M1 Pro MacBook Pro, meanwhile, only heated up to a maximum temperature of 201º Fahrenheit or 93º Celsius, and as I said before, was much quieter throughout the test. I couldn’t hear the fans at all, and iStat Menus said that they were running at idle speeds. Though the older M1 Pro might not necessarily be more power-efficient than the new model, which I’m not surprised about considering Apple touted power-efficiency as a major selling point of the new chip (see battery life portion), it runs cooler and because of that, beats the M2 Pro when it thermal throttles in *some *scenarios (I’ll get to that a little later). The last-generation MacBook Pro also feels more pleasant to use due to its cooler running temperature and lower fan speeds, which is just another blow to the otherwise great M2 Pro.
I also had to test my M1 Max Mac Studio, and of course, it was the coolest computer of the three. The Mac Studio, with its chronically superior thermal system, reached a maximum temperature of only 143º Fahrenheit, or 61º Celsius while remaining at its idle fan speed throughout the test. It didn’t thermal throttle at all, it didn’t make much noise, and it crushed the test — exactly as I expected it would. All this data comes back around to support my subtitle at the top of this review: the M2 Pro MacBook Pro is thermally and power-wise compromised due to its chassis. It’s not because the M2 Pro is a bad chip. It’s far from one, in fact. But I have a sneaking suspicion that the M2 Pro and M2 Max, which would thermally fair worse in this test because it’s a larger, hotter, faster chip, would do much better in a desktop chassis, and that’s going to be a theme I echo throughout this review.
All of this, whether it be battery life, efficiency, TDP, design, or anything else, comes down to performance in the end. With the M2 Pro MacBook Pro being, in essence, a minor spec bump, it’s important to see exactly how substantial that spec bump is and how the new chips fair against their predecessors and existing products they live alongside. I’m going to focus on 3 main points throughout the performance section of this review: first, I want to figure out exactly how big of a jump the new M2 Pro is compared to the M1 Pro, and in what scenarios it succeeds. Second, I want to show why synthetic benchmarks don’t show us the full picture when trying to illustrate year-over-year performance improvements between professional machines doing professional tasks. And third, I want to illustrate how the Mac Studio and Mac mini, which on paper may show themselves to be underpowered, have their place and why the MacBook Pro, considering its thermal limitations, isn’t up to par with desktop machines Apple sells. Professional work is nuanced — there are tons of different use cases for a professional machine like the MacBook Pro. While I couldn’t include a benchmark for every single use case, I was able to get a broad view of how the laptop performs in a variety of different use cases. First, I’ll start with synthetic benchmarks: Geekbench, Cinebench, Wild Life Extreme, Blender Benchmark, GFXBench, Unigine Heaven, and the Blackmagic Disk Speed Test to test how the reduction in NAND flash chips affects SSD speed. I’ll then get into actual use cases with real projects: Xcode compilation, Compressor encode, Final Cut Pro project render, and Final Cut Pro encode. This variety of tests should provide a deeper look at the CPU, GPU, storage, and media encoders of the new MacBook Pro, other Macs, and its predecessor. And let me say: the results are extremely interesting and peculiar in lots of ways. In addition to just benchmarking the 3 machines I have on me, my base model M2 Pro MacBook Pro, M1 Pro, and my base model M1 Max Mac Studio, I’ve done a lot of math to make projections based on scores from others, the scores I got, and how much of a difference there is between processor generations. Projected scores will be indicated by an asterisk next to the name of the chip on charts, and I’ll call them out when I’m making direct comparisons. With that, let’s get into these numbers.
Geekbench 5 is one of the most, if not the most popular set of benchmarks for assessing the raw performance of all types of computers. However, it’s important to note that Geekbench never pushes processors to their extremes, so they should never be used to assess how different computers may affect *your *workflow specifically, and they should never be used to assess how a machine will perform when working in certain applications. What they can do is help us figure out what Apple’s done to make the chips better, and help us figure out by how much. When performing a multi-core CPU test, the 12-core variants of the M2 line crush every other chip besides the M1 Ultra with 20 cores, which is to be expected. The 12-core chips, which I’ve projected the scores of, reach a score of 14593, more than 2000 higher than the M1’s 10-core CPU. It looks extremely impressive in these charts, and the numbers roughly match up with what Apple told us to expect during its keynote — we’re seeing about an 18% improvement year over year when we compare the 12-core scores to the 10-core M1 and the 8-core scores to the 10-core M2. Meanwhile, the 10-core chip in the M1 Max Mac Studio just barely edges out against the 10-core chip in the M2 Pro MacBook Pro due to the Mac Studio’s higher clock speed and better thermal performance — we saw this last year during my Mac Studio review. Again, these scores aren’t representative of how well the machines will actually perform in the real world, but this is confidence-inspiring. The binned M2 Pro would most likely beat the non-binned M1 Pro from just a year and a half ago. That’s very impressive!
In the Metal GPU test, we start to get a look what the M2s are capable of when stretched to their limits. And the results are… insane. More than I expected, even. The full-fledged 38-core M2 Max MacBook Pro flat out beats the 48-core M1 Ultra in my projections, with a projected score of 86021. And it even comes close to touching the 64-core M1 Ultra with its score of 88542 — that’s unbelievable for a laptop with thermal limitations. Yes, this doesn’t tell us the full story (I’ll keep saying this over and over because it’s important), but what it does do is tell us that the M2 Max is a substantial leap over many other Apple silicon chips — all but the top-of-the-line 64-core M1 Ultra. Going down the line further, We can see that the M1 Max in its top configuration in the Mac Studio, at least in my projections, still ends up winning against the M2 Max in its binned configuration, but only by a relatively small amount. The raw GPU increases are insane here with the M2 Max, more than I expected at over 21% year-over-year. I imagine the reason the Mac Studio is still able to hold a chance with the M1 Ultra is purely due to thermals and clock speed — if a 38-core SoC can get so close to a 64-core one, that has to be the reason. This doesn’t take into account any other parts of the system, but color me impressed.
Cinebench R23, unlike Geekbench, gives us a better idea about the thermal limitations of each computer. While the single-core and 10-minute tests aren’t the most useful, the 30-minute multi-threaded test really shows how these machines actually perform in the wilderness. It might not be as accurate and useful as a real project render, but it’s repeatable on every machine, and the results come close to something more realistic. Of course, the 12-core M2 Pro/Mac MacBook Pro (projected) takes the lead, but only by a couple points, which is why I haven’t really organized the bar chart the correct way with the fastest machine going at the top. The truth is, if I ran this test again, I could have gotten a test where the Mac Studio did better by 5-ish points or something. The CPU results here really are a tossup, and that’s illustrated even further when taking thermal systems into account. As I noted with the thermal section earlier, Cinebench really pushes the cooling systems of these computers to the brink of forcing the CPU to throttle, and that’s what happened here with the MacBook Pro (like I said earlier). It only makes sense that the 10-core M2 Pro should beat the 10-core M1 Max, but that didn’t happen. In fact, the M2 didn’t even come close due to its thermal constraints. See what I’m saying here? Thermals matter and I bet that if we ran this test for an hour instead of 30 minutes, the Mac Studio would have beat both MacBooks by a long shot. Putting thermals aside for a second, the year-over-year upgrades found here confirm the Geekbench results we saw earlier, with about an 18% increase across generations. That’s, once again, a big deal for people who render things for a living, and I see it as a win even though the machine throttles slightly during extended workloads.
Sticking with the 3D rendering theme, Blender Benchmark shows us a completely different story. Though the tests aren’t completely the same, they utilize the CPU cores the same way and perform the same task. In Blender, the 12-core M2 Pro (projected) crushes every other chip I put it against. It makes sense, too: the 12-core CPU in the M2 Pro has more power than the 10-core from the M1 Max, and Blender doesn’t really push the cooling system too hard for that to make a difference. The point is that if you use Blender and aren’t doing that many long renders (the benchmark only lasts about 5 minutes, so it’s never enough to test thermals), the M2 Pro/Max MacBook Pro seem to be right for you. The year-over-year improvements match up with what we’ve seen from other tests, too — there’s roughly a 20% increase in CPU performance here, with the M2 Pro/Max sampling the classroom render 52 times a minute, the junkshop render 68 times, and the monster render 117 times. Once again, this is a win.
But the real win is when Blender is set to use the GPU instead of the CPU. Apple did something extraordinary this time around, because GPU-based rendering took a humungous leap this year. Here, the M1s pale in comparison to the new M2s — both M2 Max chips rip the M1 Max into pieces, with nearly an astonishing 50% boost in performance at 780 samples per minute in the monster benchmark, according to my projections. Even though I haven’t included them in the graph (because I didn’t do all the math), I predict that the M2 Max would beat even the high-end M1 Ultra purely due to this astronomical leap in performance. Many have said that the M2 line isn’t much of an upgrade, and while that may be true in some regards (we’ll get to this later), the M2s seem to do a crazy job at GPU-based 3D rendering. Sure, the Mac Studio may perform slightly better if the renders were longer and more complex, but I don’t think there’s any possibility that any M1 chip beats its M2 counterpart, regardless of the machine’s thermal envelope and cooling system. This really knocked my socks off — I couldn’t believe my eyes. Nice work, Apple!
Continuing with the GPU theme, Wild Life Extreme is a benchmark that aims to simulate gameplay using Metal. While there aren’t many AAA titles that you could play on a Mac that would actually push the GPU to its limits, Wild Life Extreme is a really good test for sussing out how the cooling system comes into play and what’s possible with the GPU in a variety of different scenarios, including game development. Here, because the test is so demanding, we start to see more of how the M2s throttle and how the Mac Studio (and potentially Mac mini) succeed. Still though, the M2 Max with 38 cores beats every other machine I projected, with a score of 13543, which is impressive. However, it doesn’t beat the M1 Max with 32-cores by a ton, which just barely trails behind with a score of 13368. It’s important to keep in mind that the M2 with 38 cores is the successor to the 32-core in the Mac Studio. I’d call this difference between the top M2 Max and M1 Max mostly a tossup — we can’t really make a concrete comparison between the two machines because they’re so close. What we can do is compare the M1 Pro to the M2 Pro, and we see sizable gains here, with the base M2 Pro getting a score of 10606 with the base M1 Pro getting a score of 8730, a 20% gain. I have reason to believe here that the thermal systems of both machines are playing significant roles in this test. It would only make sense that the M2 Max projection would be 20% faster than the M1 Max projection, but that’s not the case at all. For now, I’m confident in saying that the GPU isn’t too much faster in a more real-world test that relies on thermal performance, and I’ll continue this theme in later tests.
Similar to the Wild Life Extreme test, GFXBench is another GPU-focused benchmark that tests Metal graphics rendering and video game-focused workflows. However, much to my surprise, the results weren’t close between both benchmarks. I ran the Aztec Ruins 4K benchmark, which is one of the most demanding ones on the list, and the M2 Max beat all other chips by a fair margin, a much larger margin than it did in the Wild Life Extreme test. Even the 30-core M2 Max wound up beating the 32-core M1 Max by 4 frames a second in my projections, which granted, isn’t a lot, but enough to draw some important conclusions. Apple has put serious effort behind the graphics performance of these new chips, and while this isn’t the most realistic test (I mean, who games on a Mac?), it sets a benchmark (🥁) for what we should expect in Final Cut Pro and Compressor (spoiler alert, it doesn’t match up with what we see here). The year-over-year leaps are substantial in this test, though: we’re seeing a 25% gain from the last generation, which is nothing to scoff at and matches up with what we’ve seen in other synthetic benchmarks.
To round out GPU testing, I ran the Unigine Heaven Benchmark, a less-rigorous, gaming-oriented graphics test that puts more emphasis on the processors’ raw power, similar to Geekbench. This benchmark, unlike GFXBench and Geekbench however, uses OpenCL instead of Metal, which means it doesn’t take advantage of the chips as much as those other benchmarks (hence why they don’t heat up during this test). Still though, many game development engines and programs on the Mac use OpenCL and Vulkan, two APIs that Macs aren’t optimized for, so this test is important. With that being said, I still would take the results from this benchmark with a grain of salt for the most part unless you know you’re directly impacted by its results. Regardless, the M2 Max AND, most surprisingly, the M2 Pro both beat the top-of-the-line 32-core M1 Max in the Mac Studio in my projections, with a score of 11459 on the Max running at 453 FPS. And to add even more insult to injury for the older chips, the 16-core M2 Pro also beats the 24-core M1 Max by a fair margin, about 26%. That’s insanity! The year-over-year improvements are even crazier, we’re talking 30% gains between Pro models. These benchmarks, as I said earlier, should be taken with a grain of salt, but the results are impressive nonetheless. They make the chips seem much more than a minor spec bump. Again, nice work Apple.
The final synthetic benchmark I did was one that has been stirring up a lot of controversy across the internet recently, and it’s regarding SSD speed. If you’ve been living under a rock recently for some reason, let me briefly explain what’s going on: early teardowns and benchmarks of the base model 14” MacBook Pro and the base model M2/Pro Mac mini, both with a storage capacity of 512GB, have been found to only have one NAND flash storage chip, containing the full 512 GB of storage. Previous models from 2021 and 2020 respectively have had 2 separate 256 GB flash chips that would be recognized as one by the system, making the total storage amount to 512 GB — this is still what the Mac Studio and higher-end configurations of the MacBook Pro and Mac mini have. The problem is that if you remove NAND flash chips, the speed of the SSD that the system recognizes as whole is lessened proportional to the amount of chips removed, which in our case is 1 of 2, which makes the new Macs have an SSD that’s 50% slower than the previous generations or higher-end configurations of the same computer. Many people in the tech space have discounted this change, saying the media has blown the issue way out of proportion to write sensationalist headlines for the sake of clicks, and to them I say, you’re probably discounting the issue because you probably have 4 TB in your machine or something. I, like many others, choose to buy the base storage amount because it encourages me to be redundant with my storage backups; I’m a firm believer in having a small amount of internal space, using it for what you’re doing immediately, and offloading the rest of your data to either the cloud or an external disk. This problem that Apple’s created as the result of a very obvious cost-saving measure makes that process of offloading files infinitely harder, since backing up or retrieving information from a disk now takes much longer than it used to. I’m incredibly against this change, and I feel like Apple needs to either right their wrong in next-generation models, or at least, at the bare minimum, tell consumers that the storage is literally half the speed of the older models before they buy the machine. This is a downgrade, and a large one at that.
The Blackmagic Disk Speed Test proves that the speed has been halved this year. As you can see from the charts, my Mac Studio (weirdly enough), takes the lead in both read and write speeds, and the last-generation MacBook Pro trails closely behind. The new MacBook Pro, however, has speeds 50% slower than the previous generation and the Mac Studio, and this is comical. It’s so low, that the write speed, which is usually slightly slower in most SSDs, is faster than the read speed. I’m really not impressed by this change; it’s backwards, a downgrade, and should be rectified. I let it slide in the M2 MacBook Air since that wasn’t a pro product where people wouldn’t transfer Xcode project files or large video files between machines or external storage, but this is downright laughable. It’s a no for me.
As I said before, synthetic benchmarks only tell us part of the story when testing out a new machine. Most of the important information comes when we actually put the machines to their paces and do some real work on them. After all, that’s what they’re for; that’s why they have the word “Pro” in them. I tested the laptop in 3 different apps: Xcode, Final Cut Pro, and Compressor, and the results were very peculiar.
Starting with Xcode, we see some interesting results when compiling my app, Citations, which is built fully in SwiftUI with 20 Swift files. It’s not a very large project, nor is it very resource intensive, but still, the Mac Studio prevails over the projections for the 12-core M2 Pro, as well as the 10-core M2 Pro by about 40% in the case of the former machine when looking at clean build times. The cached build times are a little tighter, but that’s to be expected when the build times are so short. We’re also seeing year-over-year performance increases of about 15% between the base model M1 Pro and M2 Pro, just slightly below the 20% increases Apple promised us in Xcode during their keynote. Obviously, these build times will vary between projects and simulated devices, etc., but these are sizable improvements, especially if you’re dealing with a much larger project. 15% is nothing to scoff at. Still though, the Mac Studio crushes the “more powerful” M2 Pro with 12 cores, which shouldn’t be the case according to Apple’s benchmarks. I’m thinking this has something to do with clock speed, and that the 16” MacBook Pro would perform slightly better with High Power Mode turned on, but I haven’t seen any reports of this on the internet. This is just another example of how synthetic benchmarks don’t paint the full picture; if you just looked at Geekbench scores, you would have thought that the MacBook Pros with M2 would be astronomically faster, when in reality that was far from the truth. If you’re a developer coming from an M1 laptop, you’re going to love the quicker compile times, but I wouldn’t be too disappointed if you’re not upgrading or are on a Mac Studio.
My least favorite test to perform every year is the Compressor benchmark, where I take a bunch of long 4K 30 FPS video files, which add up to be 1 hour of video, and convert them first to ProRes 422 from HEVC, then from ProRes to H.264. I started doing this test in 2021 when the M1 Pro first came out because I wanted to see how the video encode/decode engines would perform, and I still use the test to this day to get an accurate perspective of how the engines actually help with video-related tasks. This is always one of the weirdest tests I do because it’s never proportional to how powerful the chips actually are, and this time was no different. Of all computers, the M1 Pro with 14-cores flat out beat both the new 16-core M2 MacBook Pro as well as the 24-core Mac Studio in the first test, while only beating the M2 in the second test. Sure, I bet that if I repeated the test again, and again, and again, I’d maybe get the M2 to win, but this just shows how the GPU cores in an Apple silicon SoC really have no correlation to how much video you can process in a given amount of time. The encode engines in the M1 and M2 are identical, which means the changes have to come down to disk speed and thermal performance. I wouldn’t take this as concrete evidence of the M2 being worse for video work (you’ll see that isn’t true in just a second), but I am slightly concerned that the chip is overheating. Meanwhile, I have no idea what’s going on with the Mac Studio in the first test. I’d just pawn it off as an oddity.
Onto perhaps the strangest test of the whole bunch — the Final Cut Pro project render. This test is mostly symbolic and redundant, knowing that Final Cut Pro renders in the background, but it’s always interesting to see how those video engines are put to use by the system, and how fast the chip actually is when put under extreme stress. What makes this occurrence of the test so strange, however, is the fact that the 24-core Mac Studio lost to the 16-core M2 Pro by just 1 second. That’s right, one singular second. Once again, I have no idea what’s happening to the Mac Studio — it’s supposed to be 2x faster than the Pro models, but it seems like rendering utilizes more of the GPU than the video accelerators, making the M2 Pro and M1 Pro butt heads with each other. I’d imagine that if I repeated the test with an M2 Max, the machine would throttle and end up taking about the same time as the M2 Pro and M2 Max, but I unfortunately couldn’t test this because I didn’t have a machine with an M2 Max. Interesting, though.
The last test is, in my opinion, the most important test: Final Cut Pro project export. This test strains the machines unlike any other test we’ve done in the most real-world way possible. Whatever results we see here tell us what to expect out of these machines GPU-wise in pretty much any task and… wait what? Yes, your eyes aren’t deceiving you; the M2 Pro is about 20% slower than the M2 Pro, and the M2 Max beats the M2 Pro by about 50%. Once again, I’m afraid to report that the M2 Pro MacBook throttled a couple minutes into the test. I kept a close eye on iStat Menus, an app that lets you monitor temperatures and SoC utilization throughout the test, and the machine just gave up after a while before resuming the task a minute later. This is deeply disappointing and quite concerning — I assume this issue wouldn’t occur on the 16-inch model based on others’ testing, but regardless, this is a bad look for Apple. I assure you I’m not lying — the M2 Pro is noticeably slower than the last generation model, even after running the test multiple times. Perhaps the system simply isn’t optimized yet? We’ll have to wait and see, and I’ll report back with the results if a software update comes out, but for now, this is what we’ve got. If you’re a video editor, Mac Studio is still the way to go for you, but at this rate, I’m having a tough time recommending the M2 Pro in its MacBook Pro form. Here’s hoping that this is just a simple software glitch that’ll be ironed out with time.
Coming back around to the original point I made when I started this review more than 6,500 words ago: the M2 Pro MacBook Pro is conflicting. It’s an amazing laptop that grows atop the amazing platform constructed last year by the M1 Pro version, gains astonishing capabilities that blow me away, but also throttles and falls flat on its face when compared to its predecessor in some rare cases. The 2023 MacBook Pro is a great laptop — I can’t overstate that. But do I think you should upgrade, or even more interestingly, ditch your Mac Studio for it? Absolutely not. Yes, it’s 20% faster in nearly all synthetic benchmarks. Yes, it lasts longer. But it also heats up tremendously, throttles like no other Apple silicon laptop, and has a worse SSD in it. If this is your first Apple silicon laptop, or if you’re upgrading from a base M2 or M1, then I can strongly recommend the M2 model as much as I did the M1. But chances are, most of you reading this aren’t in that situation, and for you, I think it’s safe to say that this is simply another spec bump worth skipping.
Where the (otherwise fantastic) value proposition of this laptop comes into jeopardy is if you can find a good deal on an M1 Pro that’s refurbished. I don’t think the upgrade is worth $400-$500 for most people. But if this is your only choice or you want the latest and greatest, you won’t regret the purchase one bit. I know I certainly won’t. It’s not perfect, but damn, does it ever come close.
This review was produced following 1 week with the MacBook Pro with M2 Pro processor, 512 GB of storage, 16 GB of RAM, 10 CPU cores, and 16 GPU cores. This unit was purchased by Mac Technophilia via the consumer Apple Online Store, and was not provided as a review unit. Apple had no editorial input into the making of this review, and they’re seeing it for the first time you are. This article has no sponsors, no ads, and no paid endorsement of any brand whatsoever — the words you read are mine and mine only and aren’t influenced by any third party with large pockets. If you enjoy that type of journalism, I encourage you to follow on Medium or add Mac Technophilia to your RSS reader for more educational, in-depth, and nerdy content like the article you just read. Trust me, there’s tons more coming, and you won’t want to miss it. Be sure to mask up, stay safe out there, and stay healthy — and stay curious.
