The increase over M3 is almost entirely due to increase in clock speed, and the effect of Geekbench 6.3 supporting SME for a small subset of tests, which leads to those tests running ~2x faster on M4. The actual IPC increase on general purpose code is small, around 3%. Apparently Apple has added SME to the M4 cores, which is cool, but makes comparison to previous cores a bit suspect unless your specific application can actually benefit from SME. Source: https://forums.anandtech.com/threads/apple-silicon-soc-threa...
1. Apple changed the design of its cores to achieve higher clock speeds without raising power. In other words, they deliberately decided to alter the design to get these results.
