Odd cores, AV1, and other things to know about the M2 Pro and M2 Max

The march of Apple silicon is not slowing down. To power the new high-end Mac mini and 14-inch and 16-inch MacBook Pro systems, Apple has unveiled the all-new M2 Pro and M2 Max processors, taking all the improvements of the A16 processor and applying them to the Mac chips, resulting in a significant boost in performance over the M1 line.

Apple claims the M2 Pro and M2 Max have 20 percent better CPU performance than the M1 Pro/Max, and 30 percent faster GPU performance. They also include the newer Neural Engine, which is up to 40 percent faster than that in the M1 series.

As impressive as these new chips are, they are unsurprising. Some of the rumors about more advanced manufacturing, new GPUs, or other improvements have proven untrue, and the M2 Pro and M2 Max appear to be exactly what we expected: the M2 scaled up with more CPU and GPU cores and a wider memory bus. In fact, the M1 Pro/Max chips had more to differentiate them from the M1 than the M2 Pro/Max do from the M2. Here are some of the key characteristics worth noting.

Same CPU core design and performance as M2

When Apple introduced the M2, it claimed an 18 percent improvement in CPU performance versus the M1. Now it’s saying that the M2 Pro and M2 Max have 20 percent better CPU performance than the M1 Pro and M2 Max. In other words, you can expect the same overall architectural improvements and performance uplift as the M2 delivered compared to the M1.

Apple

Some of this simply comes down to core count. The M2 Pro and M2 Max have eight high-performance cores just like the M1 Pro/Max, but the number of efficiency cores goes up from two to four. So the M2 Pro and M2 Max have 12-core CPUs while the M1 Pro/Max have 10-core CPUs.

GPU gains are better, but there’s no great architectural shift

Apple says the GPUs in the M2 Pro/Max are 30 percent faster than in the M1 Pro/Max. That’s unsurprising, considering the M2 was said to have graphics performance that is up to 35 percent greater than the M1.

This is likely due to the addition of a few more graphics cores and some minor architectural tweaks. The M2 Pro has up to 19 graphics cores (vs. 16 in the M1 Pro) and the M2 Max has up to 38 graphics cores (vs. 32 in the M1 Max).

Those are unusual numbers—core counts are typically even in number, and often powers of two, so it’s likely that the design is for 20/40 GPU cores with room for one faulty core to be disabled, improving manufacturing yields on these big chips. Still, that’s a nearly 20 percent increase in core count, so with some improved caching and higher clock speeds, you can get to 30 percent.

Apple

If you were hoping for an all-new GPU from Apple with modern features like ray-tracing acceleration, which GPUs from AMD and Nvidia have had for several years already, you’re going to be disappointed.

Memory bandwidth is the same even with more RAM

Memory bandwidth on the M2 Pro is 200 GB/sec, just as the M1 Pro. The M2 Max doubles that to 400 GB/sec. So we’re almost certainly still looking at a 256-bit LPDDR5 memory bus on the Pro and 512-bit LPDDR5 on the Max. Apple improved caching in the A16, and those architectural improvements likely carry over to the M2 Pro and M2 Max.

While the M2 Pro maxes out at 32GB of RAM (the same as the M1 Pro), the M2 Max can go up to 96GB–50 percent more than the 64GB limit of the M1 Max. But there’s a catch—the 96GB RAM option is limited to the version of the M2 Max with 38 GPU cores. The less expensive option with 30 GPU cores tops out at 64GB.

Only one video encode engine, with no AV1 support

With the M1 Pro had a new video encoder and decoder that added hardware acceleration for the ProRes format, and the M1 Max doubled the number of encoders to two.

Oddly enough, that’s not the case with the M2 Max, which still has just one video encoder. It supports all the same formats as before: H.264, HEVC, ProRes, and ProRes RAW. That means support for the newer AV1 codec is still nowhere to be found, despite being available in today’s GPUs from Nvidia, AMD, and Intel. Apple is falling behind here.

Apple

We couldn’t help but notice that all of Apple’s marketing benchmarks around video production, claiming improvements of 30 percent or so, are for activities like color grading or 3D effects rendering, which are typically limited by GPU performance. Apple didn’t provide benchmarks showing an improvement in video encoding speed, so it will be interesting to watch the first reviews of the M2 Max to see if a single new video encoder can keep up with two encoders in the M1 Max.

It’s all still 5nm

There has been a lot of speculation about when Apple will start using TSMC’s new 3nm manufacturing process. Apple is said to be the first major manufacturer to do so, and some of the rumors said the new M2 Pro and Max chips would be where Apple makes the leap.

It looks like we’ll be waiting for an M2 Ultra later in the year, or possibly the A17 that will be introduced with the iPhone 16 this fall. Apple says the M2 Pro and Max are made using “second-generation 5nm technology.” It’s likely that this is the same process that Apple somewhat erroneously called “4nm” in its iPhone 15 presentation in September 2022. TSMC calls its N4 node an “enhanced” 5nm process, and Apple seems to be correcting its language here.