Is this useful? https://barefeats.com/quakeX.htm

Unfortunately 3D graphics performance is usually the Achilles' heel on these Macs because drivers/extensions are not as polished as in Windows.

pretty biased benchmarks and obviously when you are testing 2 different CPUs you use the same graphics card on both. But also I would of liked to see some productivity benchmarks, no one is buying a mac to game on. I saw the 2nd graphs and its beat by a P4 1.6ghz. so pretty much a 1.4 Tualatin would probably smoke it.

🤣 don't worry Steve Jobs has you covered

In the day, I recall Apple would state that the G4 was equivalent to a Pentium 3 at twice the Mhz - for example a G4 500 was equivalent to a Pentium 3 1Ghz. It was never the case of course, games performed terribly compared to Wintel equivalent, however FPU performance was strong. Mac OS X also liked dual G4 CPUs, seemingly working better and more smoothly than an equivalent Windows supporting dual CPUs of the day.

I'm pretty sure they were building worlds fastest supercomputers in the late 90s from PPC cuz they were so fast.

kolderman wrote on 2020-09-24, 05:46:

I'm pretty sure they were building worlds fastest supercomputers in the late 90s from PPC cuz they were so fast.

fastest super computer of the late 90s was built by intel, https://en.wikipedia.org/wiki/ASCI_Red

I was thinking of Blue Gene but it's a lot later than I thought, only started development in 99.

There are some comparisons here:
https://gmplib.org/gmpbench

This is a synthetic cross-platform bench which is single-threaded and integer-based. Looking at the compiler versions and flags it's not always a level playing field (some versions far more (CPU) optimized than others), but it does allow comparison of very diverse architectures. Scroll down to the bottom for 1GHz G4 vs various other things. Note that the lower clock of the G4 is to be expected - this is a snapshot in time and G4 was running against P4 and AthlonXP CPUs clocked significantly higher.

Conclusion is clear: at least in this benchmark the G4 comes near bottom. even if you correct for clock speeds. Clock-for-clock it only manages to beat the Via Nehemiah. Even the P4 Prescott, with notoriously low IPC, manages better (almost 4x the performance while clocked 3x higher).

Now that's ALU which was know to be a G4 weak spot, I'm having trouble finding good FPU data that wasn't pure marketing.

G4 were fast on specifically tailored software, which is similar to approach in gaming consoles (at least in first party titles). Like Final Cut Pro for example. So it's fair to say, that without Apple ecosystem, they were mostly crap or at least not general purpose friendly CPUs.

The Serpent Rider wrote on 2020-09-24, 12:44:

G4 were fast on specifically tailored software, which is similar to approach in gaming consoles (at least in first party titles). Like Final Cut Pro for example. So it's fair to say, that without Apple ecosystem, they were mostly crap or at least not general purpose friendly CPUs.

That was probably the Altivec instruction set. In a way that made G4 similar to P4: instead of going for good general-purpose design, Motorola went all in on software supporting their special instructions, just like Intel did with SSE2 and extremely deep pipelining on Netburst. Just as with Netburst, it wasn't a winning strategy in the long run - particularly as the G4 couldn't be ramped up to insane clock speeds to compensate.

particularly as the G4 couldn't be ramped up to insane clock speeds to compensate.

Now that you mentioned it. G5 flopped on quite spectacular level due to that. Stock water cooling and huge-ass "tower" coolers with heatpipes was just bonkers in 2005-2006. Even Pentium 4 Prescott wasn't THAT bad.

They generally lagged behind, as others have already stated. It's madness when you consider that they were still using the G4 as late as 2005 for their laptop range.

The G5 was a little more competitive against x86 for a while (pre dual-core CPUs), but severely limited by the fact it ran hot as hell, hence the long life of the G4 way past it's prime.

With regards to games, a lot of Mac ports were very poorly optimised in comparison to their PC versions. So the poor optimisation combined with the less capable CPU and you often got much worse performance overall.

For what it's worth I have both a Mid 2005 iBook G4 (1.33 GHz 7447a G4, 1.5 GB DDR266, Radeon 9550 with 32MB of dedicated DDR Vram, SSD) and a late 2003ish Acer TravelMate 290 (Banias PM 1.4GHz, 2GB DDR 266, 855GM integrated graphics with shared memory, Hitachi 7K2 spinning rust.
On paper these machines should be comparable, with the iBook having the upper hand with its dedicated graphics not eating memory bandwith and solid state drive. In actual use, though, the Acer feels much more "snappier" and usable.

After upgrading it to a 2.13GHz Dothan (required a PLL mod that also clocked the RAM at DDR356), which would be a config available in a 2005 laptop anyway, it feels like a modernish, still plenty usable computer, while the iBook is just a nice to look at but annoying to use toy in comparison.