Macworld
When we reviewed the first Mac with an M5 processor—the 14-inch MacBook Pro—we were pleasantly surprised. While most of the laptop hadn’t changed much (outside of a much faster SSD), the leap from M4 to M5 resulted in a pretty big performance jump.
Despite still employing a 10-core CPU and 10-core GPU as in the M4 before it, the M5 version of the MacBook Pro saw single-core CPU gains of around 13 percent, an even bigger multi-core boost of about 22 percent, and extremely impressive graphics performance uplift of 35-50 percent.
This year, possibly in just a couple of weeks, we’ll meet the rest of the family, starting with the M5 Pro and M5 Max in the 14- and 16-inch MacBook Pros, likely followed by spring launches of the M5 Pro Mac mini and M5 Max and M5 Ultra in the Mac Studio. Here’s what we expect from those new chips.
Same core count, but big improvements
We don’t think Apple will increase the maximum number of CPU or GPU cores. Rather, we expect the M5 Pro to still have a maximum of 14 CPU cores, and the M5 Max will have a maximum of 16 cores. Some product configurations may have fewer, as Apple has always done.
The GPU, similarly, will likely have a maximum configuration of 20 cores for the M5 Pro and 40 cores for the M5 Max. We expect other particulars, such as memory bandwidth, video encoders, and so on, to follow a similar pattern as we saw with the base M5.
In other words, by looking at the leap in performance between the standard M4 and the M4 Pro/Max, and then applying that to the base M5, we can come up with a pretty good estimation of what the M5 Pro/Max performance will be.
Those are pretty stellar estimated scores for the top-end M5 Max. Around 4,500 for single-core CPU performance is a lot, but over 31,000 for the multi-core score is astounding. That’s in the same league as chart-topping 64-core AMD Threadripper CPUs.
The M5 Max may be the first Apple GPU to break 250,000 on the Geekbench 6 GPU compute test. Even the M3 Ultra, with its 80 GPU cores, landed just a hair under it. If Apple can get there with half as many GPU cores in just two processor generations, that’s an impressive pace of improvement.
The most impressive gains on the base M5 were in graphics performance, and if that holds true for the M5 Pro and Max, we’re in for a treat. A score over 2,300 on Steel Nomad is a little better than the laptop version of the GeForce RTX 4050. The bigger M5 Max, with a score over 4,600, would be comparable to a GeForce RTX 4070. Apple is still a long way away from matching the performance of the best modern desktop GPUs (such as the GeForce RTX 5080 with a score over 10,000), but this would be a big step forward.
Flexible configurations?
The M5 Pro and M5 Max may have one more chip up their sleeve: more flexible configurations. A recent rumor suggested that the M5 Pro and Max will be fabricated with separate CPU and GPU areas, linked together on the same silicon substrate. In other words, separate pieces of silicon form a single chip. This would allow Apple to freely mix and match CPU and GPU core counts.
It’s unlikely that Apple would allow any configuration you can think of, but it may be possible to buy an M5 Pro or M5 Max chip with more GPU and less CPU, if that would be a better balance for the workloads you run.
This is all thanks to the chips being manufactured using a new TSMC packaging technology called SoIC-mH (System-on-Integrated-Chip, Molding Horizontal). This may also allow Apple to integrate the RAM more tightly into the packaging, so as to improve memory bandwidth.
Flexible configurations of this sort would be new for Macs since the end of the Intel days, and there’s no guarantee it will happen. Apple could be using this new technology just to have separate silicon for the CPU and GPU in order to improve yields and heat dissipation, both of which should be improved over a single monolithic die.
We shouldn’t have long to wait to find out what Apple is up to. The M5 Pro and Max chips should debut soon, possibly later this month when Apple Creator Studio arrives on Wednesday, January 28.
