Strange CPU effect on G5

[Roman78]

Some friend asked me for two 1.8 CPU's, and i have exact 2 1.8 that are working. And a 2.0 that is also working. So i thought i put the 2.0 in my 1.8 machine. It works, it is recognized as 2.0 but i noticed while benchmarking, that the speed goes up and down.

This is a Xbench of the 1.8 and that is very constant:

This is 3 times the 2.0:

As seen it is some kind of rotating. Running something like Cinebench is slower than on the 1.8

The ASD-cd tells me that i installed a CPU whit a Faster MLB than the one on the mainboard. The 2.0 has 1000 Mhz and the mainboard has 900 Mhz. I think that there is the Problem. Is it possible to modify the Mainboard to 1000 MLB? or the CPU to 900? The Thermal Calibration worked.

 

[flyinmac]

You can run benchmarks all day long on any computer and get different results all day long.

If the average is good, or better than the 1.8 then that is what you’re looking for.

But you could also be running into an optimization issue too. Some PPC models worked best with code that was written to specifically work with its design. And performance would suffer if the code hadn’t been optimized for that newer processor.

I’d have to do some digging on the exact revision 2.0 that you have vs the benchmark you’re using.

The G5 is also a beast of heat. So it could be slowing as it gets hotter. So I’d verify your cooling setup. Good contact, not too much thermal compound, applied properly, etc.

 

[Roman78]

I just replaced the thermal compound yesterday because of failing ASD thermal Calibration.

Well it is slower whit the 2.0 in several tests. When running xbench on the 1.8 it has very constant values. It is sometime 5% up or down, but that's it. But whit the 2.0 it is totally Bollocks. Also whit cinebench 9.5: The 1.8 has a score of 250 and the 2.0 of 183. (did that test only once).

Also on my other macs is xbench very constant whit the results. Never seen anything like this before.

 

[flyinmac]

I just replaced the thermal compound yesterday because of failing ASD thermal Calibration.

Well it is slower whit the 2.0 in several tests. When running xbench on the 1.8 it has very constant values. It is sometime 5% up or down, but that's it. But whit the 2.0 it is totally Bollocks. Also whit cinebench 9.5: The 1.8 has a score of 250 and the 2.0 of 183. (did that test only once).

Also on my other macs is xbench very constant whit the results. Never seen anything like this before.

That’s right... I forgot about the old calibration stuff with the PowerPC generation.

I agree that something is definitely sounding strange.

But, I do think your original speculation might be correct.

The CPU’s performance is based partially on BUS speed.

It is set in the chip to run at “x” times the BUS speed.

So.. “x” times 900 is less than “x” times 1000.

Effectively dropping the speed of the 2.0 GHz CPU.

I’m sorry I missed that detail in your first post.

The 1.8 GHz would be a factor of 2 times 900.

2 GHz would be 2 times 1000.

If the multiplier stays at 2, and BUS speed stays at 900, then the best you can get is 1.8 GHz.

Why is performance lower then? We now need to determine voltage. Is the 2.0 being underpowered? That I’d need to look up.

There isn’t a way to just overclock the PowerPC that’s as simple as flip this switch. You can’t raise the BUS speed to 1000 without breaking the timing of everything else in the system. Because that BUS speed is determined by a multiple of other slower components speed. Adjust one, and they all get out of whack. Then your PCI slots are messed up. And your memory. Etc.

So there isn’t really a good path forward except to put it back as it was.

Edit:

Looked it up... your 1.8 has a 600 MHz BUS not 900. At least according to what I see. Could be wrong.

If true, then your 1.8 is at 3 times 600 MHz.

Taking a G5 rated for 2 times 1000 MHz, and you’re only running the G5 2.0 at a speed of 1200 MHz or 1.2 GHz.

That could be your issue.

The multiplier is locked in the CPU. So you’re not going to be able to set the CPU to a 3 times multiplier if it was meant for 2 times.

And actually what you need to get to 2 GHz is 3.3333333333.... times or 3.34 times 600 MHz.

It’s possible if you can find a software title to do it in software. But I’m not aware of any PowerPC titles to do that.

You can’t adjust your memory, your PCI Bus, etc. you’re only option is adjusting your multiplier. Which just wasn’t commonly done in the PowerPC erra.

Some computers like the mini had solder pads to adjust the multiplier. But your machine didn’t expect to have a CPU option that would require adjusting the multiplier.

That and the mini was a G4. I think all G5’s had the multiplier in the CPU. Though I could be wrong.

I haven’t done much in the PowerPC hardware for a while.

It does seem that at some point, there was speculation that Apple might have used another chip inside the Mac to prevent/lock the multiplier on the G5. But, if that is true, then your 2.0 should be trying to run at 1.8 with a 600 MHz BUS. Which it wasn’t optimized for. But could theoretically perhaps work.

I really don’t recall much success in just straight swapping G5 CPU chips.

 

[AphoticD]

I'm curious, which G5 is this? I have the PowerMac7,2 model released in late 2003 (M9393LL/A) which was a dual 1.8Ghz machine that I upgraded to Dual 2.0Ghz. ASD passed with flying colors (after a CPU repaste), the system bus speed bumped up accordingly. I didn't see any errors about it.

I wonder if you might have the 7,3 or 9,1. Does it have 4 or 8 RAM slots? Apple did a bit of a dance with the G5 series, releasing newer products which had less capability than older products for the sake of PCI compatibility, reduction in cost and requirements for thermal capacity.

 

[Roman78]

Mine should have 900 Mhz, this also tells the ASD-cd. I have the 2003 Model.

I have 8 Ram slots.

https://everymac.com/systems/apple/powermac_g5/specs/powermac_g5_1.8.html

But that is the reason why the one 1.8 i Have here won't work. I have 3 1.8 CPU's and one is not working. I thought it would be defect. But it looks different. It could be that that is a 600Mhz CPU.

The G4 were easy to overclock. Did this several time. Lately my Mini and also a Quicksilver. G3 were even simpler, they had Jumpers Damn good old times.

So the 2.0 won't operate optimal on a 1.8 Mainboard. Well i don't use it much. So i don't really care. It is more a heater than a Computer

 

[AphoticD]

The G5 will surely keep you warm during winter

They are a bit of a mystery bag and tend to have a mind of their own when it comes to fan speeds.

When you run your benchmark, have you set your Energy Saver CPU performance settings to Automatic, Reduced or Fastest? If it is on Auto, the machine may be simply throttling it's CPU(s) to help dissapate some heat.

 

[Roman78]

Does a G5 has Energy saving options? It will be on default, did not changed it after the installation. I take a look tonight.

 

[MagicBoy]

Yes. It's in the Energy Saver panel in System Preferences. Should be options for Automatic, Highest and Reduced. Automatic in theory switching between Highest and Reduced according to CPU load. By modern standards it's quite crude.

 

[pl1984]

Something seems unusual with the speeds you're reporting. I would expect a 2.0GHz CPU to operate at 1.8GHz when it is placed into a motherboard which supports the 1.8GHz clock. Typically the multiplier is set in the CPU and the clock set by the motherboard bus speed. For the model G5 you have the motherboard bus speed is 900MHz, the CPU has a 2x multiplier, giving a CPU clock speed of 1.8GHz. If you only changed the CPU with a 2.0 capable G5 with a 2x multiplier then the 2.0 capable CPU should operate at 1.8GHz.

Have you determined the actual clock speed the CPU is operating at? You mentioned ASD was reporting a 2.0 capable CPU but it's not clear to me if that means it's operating at 2.0GHz or just that it's certified 2.0 GHz CPU (i.e. by its part number).

 

[Roman78]

That's what i thought: The 2.0 will operate at 1.8.

But somehow it isn't. Xbench and "about this Mac" report that there is a 2Ghz CPU installed. ASD does not tell me that there is a 2Ghz CPU, it tells that the CPU installed has a MLB of 1000Mhz and the Logicboard has a MLB of 900Mhz. So the CPU must also have a Clock generator inside, not only a Multiplier.

 

[flyinmac]

That's what i thought: The 2.0 will operate at 1.8.

But somehow it isn't. Xbench and "about this Mac" report that there is a 2Ghz CPU installed. ASD does not tell me that there is a 2Ghz CPU, it tells that the CPU installed has a MLB of 1000Mhz and the Logicboard has a MLB of 900Mhz. So the CPU must also have a Clock generator inside, not only a Multiplier.

As I lightly touched on in my prior post, there was speculation in the past that there was a chip Apple used on either a daughter card or mainboard, that also controlled the G5 CPU speed. And that said chip was matched to the CPU installed in the system originally. Essentially a secondary control to prevent simple swapping.

Whether that was ever proven or disproven, I don’t know.

But if it was true, it could be confusing the situation, since it’s not finding it’s mate.

I don’t know if there’s any validity to that theory though, since I don’t know if anyone ever proved whether that chip actually existed in the G5 systems.