It’s really not a matter of speed. The M1 bests most Xeon and Epyc models in terms of single core performance. Heck an i9 of the same generation will best a Xeon in single core performance. The value proposition of a Xeon is in parallel performance - having dozens of cores to run dozens of threads in tandem. If rumors are to be believed then Apple’s already made test chips scaled up to this level that will likely blow Xeons out of the water performance-wise.
The Mac Pro represents a unique challenge for the M series not because of power but for other reasons. Pro users who buy this machine expect to be able to customize the hardware to their needs - adding GPUs, memory, storage, networking etc. as needed. Contrast that to the M series’s SoC design, such as memory that’s integrated directly into the package. A Mac Pro user can stuff up to 1.5 TB of RAM into their machine if they need it. Apple isn’t going to want to produce dozens of possible memory configurations for this much of a niche market, so they need to figure out a different memory model to allow for upgradable RAM. The M1’s GPU uses tile based deferred rendering, a rendering model no other GPU on the market supports. Apple needs to either figure out if they want to build dedicated GPU expansion cards that also use this model or allow the use of Immediate Mode Rendering GPUs instead.
Given how small the market is for the Mac Pro it makes sense for Apple to focus on M1X MacBook Pros, Mac Mini, and iMacs first, as this will cover the much larger prosumer and some of the professional market and be more than enough power for those users.
