This is already happening at small scale: Forum member
@herd has a thread here
https://68kmla.org/bb/index.php?thr...c-also-complete-computers-usa-shipping.38413/ where some upgrades and swaps are taking place. I believe
@LightBulbFun and
@max1zzz are also doing some, but I don't remember off the top of my brain where their threads are. (Sorry!)
There's also this discussion:
https://68kmla.org/bb/index.php?threads/so-i-bought-a-1ghz-7455.38975/ from one of the people you see in that OS9Lives thread where someone else is going down basically the same general path of putting a new CPU on an existing module or using newer CPUs to build a new module.
There's a couple other discussions,
@trag is in on a couple.
I think the real limiters are whether or not these CPUs are still being built and whether "someone" who wants to build the upgrade modules can get the right connectors and a board made, or, a steady supply of donor modules to be swapped.
Proverbially, the sky is the limit and it's super fun to see a reupped interest.
For the most part, where this is landing is re-creating what Sonnet did in early 2007 with its 1.8/2.0GHz upgrades for some G4-class machines. Prior to the launch of Adobe Creative Suite 3 (The Universal Binary version) Sonnet strenuously advertised that a 2GHz G4 could outbench a 2GHz CoreDuo in some Creative Suite 2 tasks, so, there are definitely some gainz to be made.
Doing some CPU swapping on a Mini G4 and on MDD G4 modules is probably where the best gainz are to be made, since you can start with a platform that has a 166MHz bus instead of a 100-133MHz one.
For anything older than a PowerMac G3: I think that the potential for those machines is essentially already fully tapped out. The Mach5/Kansas 8600 and 9600 have 50MHz buses and you can, nominally, put a 1GHz G4 on them. It's, if I'm remembering correctly, meaningfully slower than a 1GHz on a 100/133/167 bus. Would a PCI PowerMac with a 2GHz G4 in it be fun? Yeah absolutely. Would it be any better than a stock 1GHz QuickSilver'02? No probably not.
Not that people shouldn't pursue this, but:
The big bummer is that even on 166 bus platforms there's definitely a practical upper ceiling for how much a faster CPU will help. To be straight-up, Mac OS 9 barely takes advantage of anything more than about 300-500MHz of G3 or G4, and can not effectively use more than about 512 megs of RAM on the outer reaches of reasonability for most workflows. Mac OS X can, but, I suspect interest in faster PPC upgrades died after about 2007 because, well, a 2GHz Core2Duo should very much outperform a 2GHz G4 when both are running native code. (It's also fairly close even when running PowerPC code under Rosetta.) For OSX/PPC code you can buy a Mac Pro 5,1 or any of the early-mid 2011 Macs, all of which run OS X 10.6 and Rosetta and, to be honest I bet you could outbench any G4 or G5 at PowerPC software in that way.
Is a 2GHz G4 fun? yes.
Is a 2GHz G4 practical? maybe not -- but that doesn't mean we shouldn't do it. (And, a lot of what seems to be happening is increasing the availability of >1GHz upgrades for older machines that shipped with 400-1200MHz CPUs, and that's good for sure.)