Apple’s Mac Studio will help you keep your job . . . for now

Apple has designed the Mac Studio for professionals who have demanding workloads, that need more processing power, more graphics cores, more cooling and more memory than you can get from a MacBook or even a Mac mini.

Though, that said, you may be onto something. Having reviewed the Mac Studio with M2 Ultra here in the Digital Life Labs these past few weeks (there’s also a slower M2 Max version we didn’t review), we want one now, too.

Pro: Compared to the Mac mini and the MacBook Air, the Mac Studio offers a number of useful advantages, even for those of us who aren’t the creative types Apple always pitches its machines to, and who don’t edit videos for a living.

The Mac Studio we reviewed was much, much better at handling machine learning, a demanding workload we’re all going to have to master sooner or later if we want to stand any chance of holding onto our jobs.

For instance, we trained a BERT (Bidirectional Encoder Representations from Transformers) machine learning model on the M2 Ultra Mac Studio, and measured an inference speed of 122 samples per second, much faster than the 33 samples per second we got with the same workload on a Mac mini running an M2 Pro processor, and a little faster than the 99 samples per second we measured on last year’s M1 Ultra Mac Studio.

Con: That’s still quite slow compared to the 1200+ samples per second you can expect to get, running the same machine-learning workload on a computer with a high-end Nvidia graphics card installed.

Pro?: But the new Mac Studio is good for other stuff I need to stay ahead of the AI bots, right?

Pro: Yes.

As well as tinkering with machine learning on it, we’ve been using the M2 Ultra Mac Studio for software development, and it’s easily the best machine we’ve ever used for that.

I don’t have specific numbers to throw at you, due to the way software build times are determined by a number of factors of which processor speed is only one, but I can tell you that compared to the 15-inch MacBook Air we’ve been using as our main development laptop lately, the Mac Studio does save us many minutes a day, sitting around waiting for tests to run and software to build.

And when you’re running from robots, every second counts.

Con: Software development is getting overtaken by the bots, too.

Pro: At least, when AI comes for your job, and you’re on the bread line, the Mac Studio with M2 Ultra won’t cost a lot to operate compared to some other high-end desktop PCs. It idles at 10 watts and uses a maximum of 295 watts, according to Apple, which is a half or even a third of the power consumption of comparable Intel- or AMD-based desktop PCs with fancy graphics cards.

Con: Though, that said, a 1000-watt Windows or Linux PC would likely have better machine-learning performance, so you’d still have your job, wouldn’t you. And that Windows PC will definitely be better for gaming, which could become everyone’s occupation in the long run.

Pro: In other benchmark tests we ran, the Mac Studio with M2 Ultra was a stellar performer.

Compared to that 15-inch MacBook Air M2 I mentioned earlier, for instance, the new Mac Studio was 8 per cent faster for single-core performance (it basically has the same processor, just with more cores, so expect single-core performance to be similar), 116 per cent faster for multi-core performance, and 362 per cent faster for graphics, according to the Geekbench 6 benchmark.

Compared to a Mac mini running an M2 Pro chip, those figures are 6 per cent, 50 per cent and 155 per cent, respectively.

And compared to last year’s Mac Studio with M1 Ultra, those figures are 18 per cent (it’s a new chip with better cores), 20 per cent and 27 per cent, respectively.

Of course, benchmarks don’t paint the full picture. In our machine-learning tests, there were models that wouldn’t even train on machines such as the MacBook Air and even the Mac mini, due to their memory limitations. Which is to say, there is stuff that will run on a Studio that simply won’t run on other Macs at any speed.

And the improved cooling on the Mac Studio, compared to the Mac mini and especially compared to the MacBook Air, means that the Studio will be able to sustain its clock speed longer before the processor gets throttled to prevent overheating, further extending its performance lead for long-running, demanding workloads.

Not to mention, the Mac Studio has a much better collection of ports, especially compared to a notebook like the MacBook Air which has only two.

Con: On the other hand, the MacBook Air is much cheaper, and actually comes with a screen, a keyboard and trackpad built in, further extending its price advantage. The Mac Studio, like the Mac mini, is just a box. You need to supply all the other parts yourself.

Also, the MacBook Air has a battery, and is incredibly portable, which could make all the difference in the world when the time comes that we’re literally running from the robots.

Which, if the experts are to be believed, could be any day now.

Likes: Very quiet. Very fast. Extra memory comes in handy for machine learning tasks. Relatively low power usage.
Dislikes: Can’t be upgraded at all. Not as great at machine learning as we hoped, though that may improve as the Apple-Silicon-specific machine-learning software matures.
Price: M2 Ultra version starts at $6599. M2 Max version starts at $3299