What an interesting take.
I generally disagree though. Apple was the only place where you were going to get a Mac, but in general, TDPs scaled up on PPC almost as fast as it did on Intel (but, perhaps a few years later) and in general, Intel was winning in raw numbers, and in pulling the TDP back down in order to build reasonable laptops.
For me, it was about size and power - no Intel machine until the Core Solo / Core Duo could even come close to the performance while using 30 watts or less. The Pentium 4 was essentially the heating element in a space heater.
On the mobile side:
Mobile P4s and also Pentium Ms had been outperforming mobile G4s for a few years. Apple pretty clearly fudged the G5 benchmarks to the point where there were almost always P4s that were faster, and even if there weren't, you could look in the direction of Xeons.
At 35 watts, the Pentium 4 Mobiles in the ThinkPad T30 (2002) outperformed the G4 in the 1.67GHz PowerBook G4s (2005.) These Pentium 4Ms generally run at about 30w. There was
a few SKUs using as much as 35w\, but once you’re cooling 30w in a laptop, 35 isn’t an awful lot more.
At their wattage, the Pentium M made the damage worse because you're talking about chips that bring down the wattage to below 30 (22-25 generally) and ramp down the clockspeed massively but still outperform G4s and P4s.
On the desktop side:
Granted, the G5 is said to have used less power (generously: the original 970 is relatively low power but
a 970MP at 2GHz is pulling 100w) than those Xeons, but if that was your goal then you could go get a 1u system or a blade system with Sossaman Xeons in it and still outperform G4s. I also like
this page saying "The G5 isn't hot" and then (in 2004 when 50 watts of heat was still utterly insane to think about on the Mac side of things) shows that the 970FX at 2.5GHz can use 90-100W "max".
The last-gen mainstream P4s topped 65-86w (a previous generation had reached 89W.) The Extreme Edition chips (which are the equivalent of today's 6+core i7s and "core i9" CPUs, at ~140w)
were the only ones to go over 90w. Pentium D saw 95-130w chips, which is pretty much par for the course for gluing two 65w chips together on a single die.
——
Wattage comparisons are super interesting, but almost exclusively from an academic perspective. Perhaps the lesson here is that at the time, every chip company thought that throwing more power at the problem was the solution. The real problem was that the PC OEMs and even Apple hadn’t really had any experience with cooling hot chips. That experience (or: a few reasonable ideas tossed into the dark) ultimately came about. (Interestingly: If the original G5s were as low-wattage as I've seen, that system needs nowhere near the type of cooling it got -- perhaps Apple knew what was coming down the pike.)
From a “product” perspective it almost doesn't seem worth talking about the wattages on, say, a Pentium 4 when a G5's CPU isn’t far off, and when the top chip on the mainstream Haswell and newer platforms is 90+ watts, so there’s obviously still a certain amount of thought (Apple is also using these chips) that you can get horsepower from wattage.
I think the main thing IBM/Apple figured out before Intel is how to ramp the CPU frequency down when the system is idle to save energy and put out less heat for no good reason.
Even with as “bad” as the P4 was, I think Apple pretty clearly saw how much better Intel was doing with processors for personal computers.
I truly think the main reason Apple got away with this is because they were the only way to buy a Mac. I think that there’s things Apple could have done, possibly even blindingly obvious things (like better disks) to make the lowest end and portable products feel like better computers.