Apple calls its new M5 chip, found in the new iPad Pro and MacBook Pro, “the next big leap in AI performance for Apple Silicon.” That for Apple Silicon part at the end is key, as the AI performance of modern high-end GPUs in Windows desktops is vastly superior, not to mention the big GPUs used in servers.
Even for Apple Silicon, it’s a little bit of a stretch. The M5 is, by all appearances, just the A19 Pro made bigger. Not that we’re complaining. In our iPhone 17 Pro Max review, we showed just how much faster the A19 Pro can be than the A18 Pro from last year, and to get that sort of leap on the Mac is exactly what we expect.
Just a big A19 Pro
Apple didn’t get into specific technical details or benchmarks for the M5, but from what they have said it looks like the A19 Pro with more cores.
For example, Apple touts a “nearly 30 percent increase” in memory bandwidth, which is right in line with the increase of the A19 Pro (~76 GB/s) over the A18 Pro (~60GB/sec). That means about 153 GB/s for the M5, vs 120 GB/s in the M4. Memory bandwidth is good for everything, and especially big 3D rendering and AI tasks. This is due mainly to faster LPDDR5X clock speed.
Apple mentioned other GPU functions: The new Neural Accelerator in each GPU core is common to both A19 and A19 Pro, and is responsible for the claimed 4x increase in performance specifically for “GPU-based AI workloads,” which is separate from the Neural Engine. The Neural Engine does a lot of quick and low-power AI tasks, while the GPU is usually used for bigger and more strenuous tasks that can take several seconds or even minutes, or for AI model training. It will be a big deal when these fundamental designs are used in Apple’s server chips.
Apple also mentions third-generation ray tracing and second-gen dynamic caching, also both features of the A19 Pro’s GPU that aren’t found in the standard A19.
The M5 has four high-performance and six high-efficiency cores—not quite a doubling of the 2/4 arrangement of the A19 Pro but close to it. Given the use of the A19 Pro’s GPU design and memory speeds, I’d be shocked if this wasn’t the same CPU design as the A19 as well. The chief difference between the A19 and A19 Pro CPU seems to be larger and more efficient caches, and those larger caches are likely present here.
Keeping the lineup in sync
When Apple Silicon was new, the M and A series chips were somewhat out of sync. A new M-series chip would launch with core designs a step behind the A-series chips. Perhaps they’d have the new media encoders, or some other pieces.
But Apple seems to have its chips lined up well, now. We’ll see what the M5 Pro and M5 Max have in store for us, but it looks like the core designs for Apple’s chips have solidified into a coherent set, with performance and power use scaling up just by adding core counts or wider memory interfaces. For developers, this gives a wonderfully consistent target across the entire Apple hardware profile.
Author: Jason Cross, Senior Editor, Macworld
Jason has written about technology for more than 25 years - first in the gaming press, then focusing on enthusiast PCs and general technology. He enjoys learning how complicated technology works and explaining it in a way anyone can understand.