Apple Mac Studio (2023) review: the M2 Ultra rips

There is a new Mac Studio. It looks almost exactly like last year’s Mac Studio — so similar, in fact, that I have had to put Post-it notes on the two units we have in our office in order to tell them apart.

It remains a remarkably compact, very Apple-looking pro-level workstation. It also remains effectively not upgradable at all, which means you need to be very careful about what you select at purchase. It has basically the same ports that the last one had: a pair of USB-C ports (Thunderbolt 4 on the Ultra) and an SDXC reader in the front, and four Thunderbolt 4, two USB-A, one HDMI, a 10GB Ethernet, a 3.5mm audio jack, and the big Mickey Mousey power cable thing in the back.

There are a couple other tweaks. This year’s Studio supports Bluetooth 5.3 while last year’s was 5.0; the M1 Ultra could accommodate up to four 6K displays and one 4K display, while the M2 Ultra can take eight 4K, six 6K, or three 8K thanks to an upgraded HDMI port. The biggest difference, though, is the processor inside.

Last year’s Studio was configurable with the M1 Max or the M1 Ultra, which is essentially two M1 Maxes stapled together. This year’s Studio has been upgraded to the M2 Max and the M2 Ultra, which is two M2 Maxes stapled together. You can probably tell that these are different because the number in their name has increased by one. I’ve been testing an M2 Ultra unit of the 2023 Mac Studio with a 24-core CPU, 76-core GPU, 128GB of memory, and 4TB of storage. And test I did.

The M2 Ultra units start at $3,999, and our specific system would cost — wait for it — $6,799, so this is already a machine only accessible to a fairly specific group of customers. (Apple will give you “up to $1,500” for your now-obsolete M1 Ultra Studio in trade-in credit, should you decide the upgrade is necessary. Thanks, Apple!) This is not outrageous pricing compared to the Mac Pro (or, say, a Threadripper PC), but it is still, objectively, a lot of money.

Nevertheless, this is my first look at Apple’s new M2 Ultra chip, and it’s in a chassis poised to show it off to its absolute fullest potential. So I ran a bunch of benchmarks to see how well it performs. Spoiler: the M2 Ultra is faster than its predecessor, and I’m honestly not sure how much that matters.

First off: what exactly has changed? Both the M1 Ultra and the M2 Ultra are 5nm chips; there were rumors that this chip would see an architectural shrink, but that didn’t pan out for this generation.

Still, Apple has made some design changes. The M2 Ultra features 20 billion more transistors than the M1 Ultra did and can be specced higher, supporting 192GB of unified memory where the M1 went up to 128GB. The M2 Ultra can also be configured with a 24-core GPU and 76-Core GPU, whereas the M1 was limited to a 20-core CPU and a 64-core GPU. That’s more relevant to this review since I received the absolute top-specced M2 Ultra chip that one can get; we’re essentially seeing, all else being equal, how much additional performance those extra cores really deliver.

The biggest increase is in graphic performance; those extra cores are putting in the work. The differences in Geekbench’s GPU benchmarks, using both Metal and Open CL, were between 20 and 50 percent higher across trials than they were on the M1 Ultra. While this isn’t a system you’d want to buy primarily to game (it’s not matching what we’d expect from an RTX 4090, for example), frame rates on Shadow of the Tomb Raider were also 10 to 30 percent higher across resolutions, run at the game’s highest settings. The M2 Ultra even broke the 60fps barrier at 4K resolution, which is fun and neat.

The M2 Ultra also showed an 18ish percent increase over the M1 Ultra on the Xcode Benchmark, which measures compilation time. That kind of saved time could certainly add up for busy developers.

In fact, PugetBench for Premiere Pro was the only result here where the M2 Ultra got a lower score. I have run this test many times on both of our units, and I cannot figure out what the issue is; I have tweaked every setting I can think of to no avail. The score I’m getting does seem to be in line with other M2 Ultra scores I’m seeing in Puget’s database, so there might just be something weird going on. Regardless, the M2 machine blew the M1 out of the water on actual exports in Premiere Pro, so my inclination is that PugetBench and the M2 Ultra just aren’t getting along for whatever reason (not unheard of for brand-new chips).

The M2 Ultra showed a correspondingly smaller but not-nothing increase in CPU performance in both single-core and (obviously, as it has more cores) multicore performance. It’s worth noting that both machines’ Cinebench scores went down between the 10-minute and the 30-minute loop, but the delta between the scores remained fairly constant. That means the Studio’s cooling system isn’t having more trouble keeping the M2 Ultra’s temps in check than it was with the M1 Ultra.

Speaking of cooling: I didn’t notice any major differences in fan noise or heat between these two units. The M1 Ultra Studio has always been shockingly quiet for me, even with my ear to the case, and I didn’t hear anything substantial from the M2 Ultra version, either. I know complaints about the M1 generation Studio’s noise can be found on the internet, so I guess it depends on the… ear? Or environment?

Regardless. Benchmarks (and quiet performance) are all well and good, but what ultimately matters is how well the device performs in day-to-day work. Alex Parkin, The Verge’s art director for video, is much more qualified to speak to this than I am since he generally uses an M1 Ultra Mac Studio as his daily work machine. He kindly used the M2 Ultra machine for a morning of work after I plonked it on his desk in our office, completing tasks in After Effects, Photoshop, and Premiere Pro.

I asked Alex whether the M2 Ultra felt faster than the M1 Ultra, and he kind of shrugged. “Like, yeah,” he said eventually, after some thought.

He was able to complete a Premiere export in 10 minutes that he estimated would’ve taken him 20 on his regular machine (which is not too far off the results I saw from my own export testing). Nevertheless, he didn’t view the increase as particularly life-changing. His evaluation of the M2 Ultra Studio this year was several degrees more subdued than the awestruck reactions I got from our creators who tried the M1 Ultra Studio last year (who were used to Intel Macs or Windows PCs); he’s perfectly happy with the computer he has. That’s not terribly surprising — the big leap from Intel to Apple’s M1 platform isn’t likely to be replicated here.

The sense I get from speaking to Alex and from other professionals who use Apple’s desktop hardware is that the M1 Ultra is so fast that speed is no longer a hang-up in their workflow. The biggest bottlenecks in Alex’s current workday tend to be glitches in Premiere and other software he has to use, which is something I hear from folks in video and graphic design all the time — in the age of the Ultra, raw power is just not a limitation for him. Those are problems that Apple ultimately can’t fix.

(Alex also did not hear any bothersome fan noise from the M2 Ultra model and doesn’t generally hear it from the M1 Ultra, either.)

Don’t get me wrong: more speed is good. This is an impressive technological achievement. It improves upon the M1 Ultra. And on face, this is a great, if expensive, computer. The increased graphical power, in particular, is nothing to sneeze at. It also largely seems like it will serve its target audience in the same way last year’s Studio did, providing a similar, but slightly faster, experience and a similar physical presence, without necessarily fixing major hang-ups that might currently exist in their workflows.

I love machines that make it easier for people to do their jobs. Upgrading to this machine could give busy professionals a good chunk of their day back (those halved export times will add up if you’re exporting all day, every day) and could potentially make business sense. Still, it’s quite a cost, and I ultimately see this machine less as a temptation for M1 Ultra owners (outside of the most deep-pocketed companies) and more for those who are still hanging on to old Mac Pro configurations. Those were a large investment, but every ounce of extra performance the Studio can provide may bring that crowd closer to jumping on Apple’s silicon train.

As I noted in my review of the MacBook Pro 16 with M2 Max, we’ve reached the point with Apple’s chips where we’re not expecting annual earth-shattering leaps in performance; Mac computers, as is also the case with other sorts of computers, are just getting decently faster each cycle. We expect that the vast majority of Mac users will (correctly) wait a few years before upgrading. And this Mac Studio, as much as it is a purchase consideration for power-hungry professionals, is a public showcase for Apple’s engineering.