Just because this self-marketing genius needs to fill his regular newsletter doesn’t mean that there is a story worth reporting for MacRumors? “An M3 Mac Mini will at some point exist.” What’s the news value in that?
IF the story is true (big if...) the numbers are informative:
- cluster sizes remain the same for P and E cores. (There are reasonable arguments that
at some point these could change in either direction, eg drop down to 3 cores in a cluster and give two clusters; or grow up to 6 cores in a cluster; either way ramps the baseline to more cores, perhaps starting with 4+6 cores.
It all depends on usage patterns [which Apple knows, but we don't!].
You could even do something like a 3 core mini-cluster shares AMX and L2; and two mini-clusters form a 6-core cluster sharing a single L2 TLB+pagewalkers and LZ page compression/decompression hardware...)
- 10 GPU cores. We all expect the GPU cores to be more capable, not least in having more ray tracing hardware. There are two ways to improve GPU performance: making each GPU core smarter, and providing more cores.
We expect a decent GPU performance bump given that the A15 to A16 GPU bump was minimal (so basically we're expecting two years worth of improvement...).
But Apple is not trying to achieve that improvement with even a slight (eg 11 or 12 core) increase in core count; it will all be in making the GPU cores smarter.
(Which I approve of! There's definitely scope for adding more smarts to the Apple GPU core rather than just simple-minded increasing core numbers and frequency; and more smarts will probably translate into better performance for general GPGPU [including ray tracing and NERF], not just traditional triangle-based graphics.)
