We already know how well Apple's M4 chip performs


While new iPads are always exciting, the iPad Pro's M4 chip stole the show this week. The M3 just came out in October, and Apple has decided to launch its successor now on the iPad of all devices. Still, the chip is here and will likely power Apple's big AI plans this year.

Apple touts the M4's 3nm architecture, four performance cores and upgraded six efficiency cores, and claims its CPU performance is 1.5 times faster than the M2. (The company does prefer to compare its chips to the previous two generations, rather than one). Of course, the iPad isn't out yet, so reviewers and users can't test these claims. We can trust what Apple says, and we can also refer to Geekbench.

On Wednesday, M4 stats started appearing on the popular benchmarking website Geekbench. It's unclear who is testing the M4 via Geekbench, but it provides an interesting window into the chip's performance. As of this writing, single-core performance ranges from 3,595 to 3,824, and multi-core performance ranges from 13,910 to as high as 14,748 (single-core refers to testing one core of the chip, multi-core refers to testing multiple cores simultaneously). The M4 has a total of 10 CPU cores).

The iPad Pro tested was running the yet-to-be-released iOS 17.5 system, with a base clock frequency of 4.4 GHz. They come with 16GB of RAM (although it's listed as 15.05GB on Geekbench).

These results are impressive, especially considering they're running on an iPad (and the thinnest iPad ever). Looking at these scores, the M4 is slightly better than the M3. If you look at Geekbench's benchmark scores for the M3 MacBook Pro (note, not the M3 Pro or M3 Max), you'll find that its single-core score is 3,079 points and its multi-core score is 11,524 points. If you look at the sixth-generation iPad Pro running an M2 chip, the gap is even more stark: Geekbench shows the tablet has a single-core score of 2,613 points and a multi-core score of 10,082 points. Performance has improved by nearly one and a half times, which is close to what Apple claims.

I'd be interested to see how Apple performs when putting the M4 into a Mac. It's possible that Apple overclocked the M4 on the iPad because the tablet is so thin and light that it can't effectively move the heat generated by pushing the chip. In theory, even a fanless MacBook Air would have a better cooling chassis, so the M4 should perform better. If you put it in a case with a dedicated fan like a MacBook Pro, it should work even better, at least when you actually start using it.

Anyway, my main focus right now is the NPU (Neural Processing Unit) of the M4. This part of the chip is responsible for processing artificial intelligence, and Apple is likely to vigorously develop artificial intelligence after WWDC. The company says the M4's Neural Engine is capable of 38 trillion operations, which is 60 times faster than the first NPU they built for the A11 Bionic chip. (Apple again compared the statistics to the chips they released for the iPhone 8, 8 Plus, and X in 2017). While I'm sure the M4 will be the best way to experience all the AI ​​capabilities Apple is developing, I doubt it will be a huge leap over the M3, M2, or even the M1. There may be M4-exclusive features, but my guess is that Apple's entire lineup of silicon chips will be enough to run most of the new AI features in iOS 18 and macOS 15.