Geekbenching a lot of G4s

[ziggy29]

I didn't have much else to do today, so I decided to rig up all my five PPC G4 Macs to see how they compared on Geekbench 2. The baseline is a score of 1000 for the entry level G5 introduced in 2003.

I have five G4 rigs, and I just ran them through the motions. I used 2.2.0 on Tiger for all but one of them, because it only has Leopard installed (I used 2.2.7 on that one.) I was curious to see how they would all do. So from high score to low score, here we go:

1. Digital Audio, 1.467 GHz OWC Mercury G4 CPU (upgraded from 667 MHz), Radeon X800 GPU (I think it is a FireGL X3 flashed to X800 for Mac), 256 MB VRAM (with pins 3 and 11 taped to work in this model), 1.5 GB of RAM, with SSD (2001). Score: 838. Would love to see how a dual 1.8 might change the score here, but for now this will do. Also -- it seems like 2 of my three 512MB RAM modules are acting as 100 MHz RAM even though all of them are supposed to be 133 MHz (and I have confirmed it's the RAM, not the slot). I will rerun if I can get all 133 MHz RAM to show up in the System Profiler. This is the one I (until very recently) had horrible problems with heat until I installed a very loud main exhaust fan last week. I ran this test with the side door closed!

2. 12" Aluminum PBG4, 1.33 GHz, 1.25 GB RAM, 64 MB GeForce 5200 Go, 60 GB HDD (2004). Score: 792. This is the only one of these which has not been upgraded with some sort of solid state solution (usually an mSATA drive with 2.5" IDE adapter). This model is not easy to upgrade the storage so I've never bothered. I was actually surprised at how close it came to the performance of the Mac above it. But they both do have a 133 MHz bus and the newest tech built in among the lot.

3. Titanium PBG4, 1 GHz, 1 GB RAM, 250 GB SSD (mSATA to IDE), 64 MB VRAM Mobility Radeon 9000 (2002). Score: 624. Although the Pismo is my favorite Mac ever in terms of engineering (with one exception, see below), this one is my go to portable Mac. It natively supports OS 9 and Leopard and handles all of it very well. Sure, if I ever get a 17" DLSD I might go to that one, but until then....

4. Power Mac 7600, 1 GHz Sonnet Crescendo G4, 1 GB RAM, PCI Radeon 9200 graphics (128 MB VRAM) with mSATA to IDE SSD through a Sonnet Tempo PCI card (1996). Score: 435. This one was mostly held back by the 50 MHz system bus and the PCI-based GPU (the three above have AGP-based GPUs). This one is about as maxed out as you can get, and it's still marginally usable for most things 21 years later, which is remarkable in and of itself.

5. Powerbook "G3" (Pismo), upgraded with G4/550 card, 1 GB RAM, 8 MB Rage Mobility 128 GPU, mSATA to IDE SSD (2000). Score: 319. I think the graphics clearly held this one back. I love the Pismo even today as much as any Mac ever built, but the terrible GPU that can't be upgraded is a massive bottleneck here. This Mac has a 100 MHz bus compared to 50 MHz to the souped-up 7600 above, but the graphics are like an anchor strapped to its back. That said, for relatively moderate usage and even some web browsing with TFF optimized with the settings described in this forum, it's not horrible.

 

[AmazingHenry]

Pretty cool! The difference between the TiBook and the 7600 is surprising. Similar specs but different scores.

Someday I'd like to Geekbench all of my Macs, looks interesting. I've never really played with Geekbench that much before.

 

[ziggy29]

I'm pretty sure that the bus speed (50 MHz versus 100 MHz) and the 4x AGP graphics on the TiBook are why the TiBook is so much faster overall than the 7600. As might be expected the raw CPU scores were quite close (both have a 1 GHz G4).

 

[flyrod]

Similar specs but different scores.

I think that's the hallmark of this "benchmark." If you want a different score, just run it again. Here's their database results for a single machine:

Sorry to be critical, but imagine trying to build a house with a ruler that measured this well.

 

[MagicBoy]

I'm pretty sure that the bus speed (50 MHz versus 100 MHz) and the 4x AGP graphics on the TiBook are why the TiBook is so much faster overall than the 7600. As might be expected the raw CPU scores were quite close (both have a 1 GHz G4).

The Sonnet upgrade got L3 cache?

 

[ziggy29]

I think that's the hallmark of this "benchmark." If you want a different score, just run it again. Here's their database results for a single machine.

It depends a lot on what else you are doing, and have been doing, on the machine. IMO it's best done right after booting up, and before doing anything else. That is the condition under which I ran my tests. When I do that I can and do get small variances, but nothing major. If you run it under load, with a lot of other things also running at the time, sure the score is going to tank. If you run it with a lot of other processes running because of other apps you had been running before, again, you'll see a bit of a hit. Not to mention how upgradable and customizable these old Macs were, the same base model can have radically different hardware under the hood -- CPU upgrades, GPU upgrades, SSDs, different RAM footprints, et cetera. When tested as I did it, right after clean reboot and nothing else running, it's at least a good order-of-magnitude result.

In any event it was a "just for fun" thing more than an attempt to establish firm benchmarks (along the line of then OpenMark thread).

The Sonnet upgrade got L3 cache?

Yes, 2 MB of L3. It's pretty amazing when you think about it; this mobo can take anything from a 100 MHz 601 (the original 7500) through a 1 GHz G4.

 

[DoFoT9]

This very much takes me back to a simpler time. G3 is still sitting at home collecting dust.

Crazy to think just how far computing has come since these machines were released, let alone from before that.

 

[ziggy29]

This very much takes me back to a simpler time. G3 is still sitting at home collecting dust.

Crazy to think just how far computing has come since these machines were released, let alone from before that.

Or, in some ways, how far we have regressed. You can use a 21-year-old computer to actually (albeit slowly) run a modern web browser (i.e. TFF under Tiger), at least minus the video, because they were engineered to be expandable and upgradable (and because we have developers doing things like XPostFacto). Today's closed systems out of Cupertino won't be able to do that because you simply can't upgrade them. That said, Moore's Law is slowing down, which is why Apple is forcing desupport on some 2008-09 era Macs which are perfectly capable of running Sierra.

 

[Dronecatcher]

It depends a lot on what else you are doing, and have been doing, on the machine.

My experience with Geekbench is that you have to stick to the same release version throughout - I've seen a difference of 150 points between the last three PPC iterations of Geekbench on the same machine.
I've also used Geekbench to illustrate what I term a bad install - after encountering this once, the next time I Geekbenched it - new install of Leopard on 1.5Ghz 12" Powerbook: 183, reinstall on same machine: 853
Like you, I always benchmark after a reboot and allow the boot services to run out and have anything extraneous turned off (including Wifi).
It is misleading though and should only been used as a guide - many times I've had a super slick machine, only to benchmark it and then feel disappointed - suddenly it feels slower....

 

[LightBulbFun]

Very cool post OP

I just want to point out that Geekbench 2 does not bench mark/care about the GPU in any shape or form it only benchmarks the CPU and Memory system.

the score of the 7600 is very interesting from what i have seen 1Ghz G4s in G3 Beiges also get a score close to that

I too have geekbenched a lot of systems https://browser.geekbench.com/user/76775 you can see them here (click on the geekbench 2 tab to see all the PPC macs i have benched )

indeed I have noticed GeekBench scores can vary on the same machine depending on what version of Geekbench your using