Finer grain particulars concerning the eventual M2 Professional are pretty mild for the time being, although apparently will probably be manufactured on TSMC’s 3-nanometer node.
For reference, the bottom M2 that was launched this week is constructed on an enhanced 5nm node at TSMC. It packs over 20 billion transistors, which is a 25 p.c bump over the bottom M1 mannequin, and rocks an 8-core CPU paired up with a as much as a 10-core GPU and 16-core Neural Engine.
Here is a excessive stage comparability of M1 and M2…
- M2 (enhanced 5nm): 20 billion transistors, 8-core CPU (4+4), 8-core or 10-core GPU
- M1 Extremely (5nm): 114 billion transistors, 20-core CPU (16+4), 48-core or 64-core GPU
- M1 Max (5nm): 57 billion transistors, 8-core CPU (8+2), 24-core or 32-core GPU
- M1 Professional (5nm): 34 billion transistors, 8-core (6+2) or 10-core (8+2) CPU, 14-core or 16-core GPU
- M1 (5nm): 16 billion transistors, 8-core CPU (4+4), 7-core or 8-core GPU
The M2 retains the identical CPU core configuration because the M1, splitting them evenly between excessive efficiency and excessive effectivity cores. Nonetheless, Apple is claiming an 18 p.c efficiency uplift in multi-threaded workloads, and on the graphics facet, as much as 35 p.c higher efficiency.
There are different nuances that separate the M2 from the M1 as effectively, like a bump in max reminiscence from 16GB to 24GB, and double the reminiscence bandwidth at 100GB/s. And whereas nonetheless a 5nm half, Apple is asking it a second-generation 5nm node.
