Apple announced the whole M3 SoC family at once and upgraded its 14 and 16 inch MacBook Pro models along with the sad, neglected iMac
All Hallows’ Eve might be when the kids go out trick-or-treating, but the night before — this night — is when we all bob for Apples. In a spooky themed, online-only event, Apple CEO Tim Cook led a cavalcade of corporate VPs announcing a new family of processors named M3 and a series of Macs sporting the latest in Apple Silicon. Let’s dispense with the Halloween jokes and dive right in.
The M3 Architecture
In years past, Apple played it safe and announced the smallest and lowest-performing processor in a family first. The M1 debuted alongside the Mac mini and 13″ MacBook Pro / Air in late 2020, only to follow up with the M1 Pro and M1 Max a full year later. As the perfect gut punch to all those $2,000 Mac buyers, the M2 was announced just 8 months later with higher single threaded performance. And it took until early this year to get the enhanced versions, M2 Pro and M2 Max (along with the two M2 Maxes duct-taped together to make the M2 Ultra).
That wait for more powerful versions is true no more; the M3 comes out swinging right next to its two big siblings, named M3 Pro and M3 Max. But before we talk about the chips, let’s dig into what’s new.
Apple unsurprisingly didn’t talk a whole lot about its next-generation CPU architecture. What’s different about the M3 compared to its predecessor? It’s built on a 3nm process, and because of that, its performance cores are purportedly 15% faster than M2 and 30% faster than M1. Meanwhile, the efficiency cores are 30% faster than M2 and 50% faster than M1. But how did they get that way? Don’t know. Because of the advanced lithography, it could be a pure clock speed bump, but there might be more to it. We’ll have to wait for another day to find that out.
What we do know about M3 is that it’s got a brand new graphics pipeline with three key technologies. First is what Apple calls dynamic caching. In a traditional shared memory setup, at least on macOS, the software (presumably the compiler, in this case) determines how much memory is dedicated to upcoming tasks at compile time, not runtime. The GPU will reserve memory it doesn’t need until the most demanding task requires it, leaving it unavailable to the system. The M3 family has Dynamic Caching, wherein the SoC allocates memory in real time.
The other features supported by the M3 GPU are hardware accelerated ray tracing (which recently came to the iPhone 15 Pro family) and hardware support for mesh shaders. At this point, ray tracing needs no introduction. Mesh shading has only recently made its way into PC games, with Alan Wake 2 being the first to really utilize the technology that was introduced in DirectX 12.
Mesh shaders combine vertex and primitive shading into a single program, which should boost efficiency. At least, that’s how it works in both DX12 and Vulkan, neither of which are native on macOS, where Metal is the order of the day.
