Is the implication here that there was a quad socket G5?
The quad used two 970MPs, which was a dual-core chip. This chip was also used in the last dual-core PowerMac G5s in a single socket. (All of these machines had PCI-E slots instead of the PCI-X slots of the earlier models.)
Fault tolerance has nothing to do with efficiency lol.
Which is what I said? Your nitpicking about NOP generators verses clock scaling misses the point entirely, which was the G5 was missing consumer-friendly disaster prevention features present in chips predating it by two to four years.
And, frankly, maybe it doesn't matter, because the G5 was only intended to be used by a *single* customer which exclusively built highly engineered sealed boxes with no user-serviceable parts inside, so IBM could leave it up to the integrator to protect the chip from catastrophic accidental damage with additional hardware.
The point of this thread was asking if it was safe to run a G5 with parts of that safety net disabled.
The G5 also most definitely could have been implemented in a laptop, it was entirely Jobs's fault it never happened. The 970FX pulled 11W at 1 GHz and 48W at 2 GHz, which is almost HALF of the Willamette core at the same speed.
So... are you being intentionally misleading in what you're choosing to compare to what, or is it just happening accidentally? Willamette is Pentium 4 circa
2000. The 2002 Northwood CPUs used in early Pentium 4M laptops like the T30 have
TDPs of 35 watts at 2.2Ghz, and they go as low as
20.8W TDP for the 1.40Ghz model. And, let it be noted, all
Pentium 4Ms have Enhanced Speedstep; it is, alas, very true that Intel used to screw you *very hard* if you bought the Celeron version.
You never buy the Celeron version. Ever.
Anyway, even pointing out that Intel had lower power draw than the 970FX two years earlier still misses the even *bigger* point that in the laptop market Apple
wasn't competing with 2002 Northwood CPUs with 35 watt TDPs, it was up against Dothan CPUs
with a 27 watt TDP at 2ghz that, real world, could run on about
7.5 watts at ~1.5Ghz. A Dothan at 1.5Ghz will run rings around a PPC970 at 1Ghz at just about everything, and there actually remains another interesting question which is: what supporting chipset would this theoretical two-years obsolete-out-of the-gate G5 laptop even use? Apple never completed a northbridge suitable for a laptop; the U3 used in all the non-MP desktops was a toasty little firecracker that barely worked in the iMacs, let alone even a laptop as chonky as a late 2002 Pentium M machine.
So... yeah. I get it, if Apple were Alienware and didn't mind releasing a completely laughable product they could have shoved the logic board from an iMac G5 into a case with a handle and called it a laptop, but as much as I hate to admit it Steve Jobs called this one right.
FWIW, I have one of the very last gen 1.67Ghz G4 Powerbooks, with the DDR2 RAM, higher res screen, etc; a pretty beautiful machine but an absolute doorstop. Nonetheless it's pretty entertaining to note that according to
EveryMac.com its Geekbench 2 score of 843 is only about 20% slower
than the fastest score they have for a 2Ghz iMac G5. What can we take away from this? It seems to me if you do the math a PPC 970FX laptop that's running its CPU at a wattage in the ballpark of that 2002 vintage brick of an IBM laptop is going to be running substantially slower than that G4's 843 Geekbench score. Or heck, let's be super optimistic and say we get it to run a little faster than that;
maybe through some absolute miracle we're able to get it up to 1.8Ghz, giving it a Geekbench score of 1039. Unfortunately we still need to deal with the fact that a
Thinkpad T42 that's actually lighter, and arguably sexier, than the G4 laptop, let alone this hypothetical G5 monstrosity, is scoring 1172 with the near base configuration 1.7Ghz Dothan.
In short Apple was completely screwed as long as the G5, as it actually existed, was the only tool in their toolbox and no amount of nitpicking about the Pentium 4 changes that. Maybe P. A. Semiconductor's PWRficient CPUs might have miraculously saved the day but Apple would have been stuck with nothing to offer for at least another year and a half. I guess we will never know if P.A. Semi's CPU was even worth waiting for, given so far as I'm aware the only machine it ever shipped in was a piddling number of Amiga X1000s, and just *try* to get benchmarks that translate in any comprehensible way to the rest of the world out of an Amiga fan, especially one that spent thousands of dollars on a thing they really want to believe in.
