Calibration disk 2.6.3 for Quad G5?

[ziggy29]

Where can I find one of these? I've got the G5 to the point where it boots up reliably BUT CPU B quickly ramps up in temp according to the sensors and the "overtemp" light comes on (CPU A is fine). And sometimes it goes to sleep, and sometimes it crashes. It won't reboot until the sensor thinks B is cool enough to not trigger the "overtemp" indicator. But neither CPU feels hot at all, and the exhaust from the fans (stuck on max speed, 3100-3600 RPM) is not warm at all.

Before I tear this up and either attempt to fix the cooling on CPU B or replace the CPU/LCS core, I'd like to try a calibration first. For the quad my understanding is that I need the 2.6.3 version of the ASD, and I need to boot from Open Firmware. But where can I find this beast? Unlike a lot of the old abandoned software I can't find a copy for this anywhere.

Any ideas? Failing that, is there a way I can temporarily "disable" CPU B so I can perform other tasks and tests with just the working CPU? I'm new to G5s so as much as I know about PPC Macs in general, I'm kind of a noob here.

 

[Dronecatcher]

http://www.dreamsupport.us/downloads/apple-mac/Apple_Service_Diagnostic_Disks_25_GB/

 

[AphoticD]

Where can I find one of these? I've got the G5 to the point where it boots up reliably BUT CPU B quickly ramps up in temp according to the sensors and the "overtemp" light comes on (CPU A is fine). And sometimes it goes to sleep, and sometimes it crashes. It won't reboot until the sensor thinks B is cool enough to not trigger the "overtemp" indicator. But neither CPU feels hot at all, and the exhaust from the fans (stuck on max speed, 3100-3600 RPM) is not warm at all.

Before I tear this up and either attempt to fix the cooling on CPU B or replace the CPU/LCS core, I'd like to try a calibration first. For the quad my understanding is that I need the 2.6.3 version of the ASD, and I need to boot from Open Firmware. But where can I find this beast? Unlike a lot of the old abandoned software I can't find a copy for this anywhere.

Any ideas? Failing that, is there a way I can temporarily "disable" CPU B so I can perform other tasks and tests with just the working CPU? I'm new to G5s so as much as I know about PPC Macs in general, I'm kind of a noob here.

If you install Xcode v3.1.4 (on Leopard) or Xcode v2.5 (on Tiger), you can then navigate into /Developer/Extras/PreferencePanes/ and double click Processor.prefPane to install additional Processor controls into System Preferences.

You can then uncheck the CPUs you don't want to to use. Also, be sure to set the Processor Performance to "Reduced" in the Energy Saver System Prefs. This will typically make for a huge drop in running temps on any G5.

Based on your machine's behaviour, I can almost guarantee that the ASD 2.6.3 disc will report a failure during testing or re-calibration. You will want to overhaul your Quad's Liquid Cooling System with new coolant and possibly new tubing, clamps and o'rings. If that is too much of an ordeal, even just re-pasting the CPUs alone will make a big difference.

For a quality thermal paste on the G5s, I would recommend the Grizzly Kryonaut, which has a high rating for a non-conductive paste at 12.5w/mk. (Maybe other regulars here have different/better/cheaper thermal paste recommendations?)

Good luck with your Quad and be sure to let us know how you go with the repairs.

 

[ziggy29]

Thanks, both of you. I've already changed the performance to "reduced". That seems to have dropped the core temps during operation by about 15ºC each, give or take. Under moderate load CPU B is staying in the 60s that way (CPU A is was mostly in the 30s). When the machine has been off for a while and the innards are back to ambient temperature, it will usually stay on and working. But any reboot will fail until the machine has cooled down again.

I actually have another (I think) working quad CPU/LCS core I may try swapping in also. I just want to rule everything else out before I start a teardown; I suspect the calibration may fail, too. But I'll still try it first. I've never attempted it on a G5 so I'm a little gun shy, but it's (largely) a 50-pound brick if I don't. After that I'll do the thermal paste and try again, then I may swap out and see if my other core is better. If so I'll have to overhaul the LCS, probably.

[Edit to add -- got the Processor preference loaded -- I assume switching to "dual core" (using only CPU 1 and CPU 2) corresponds to CPU A's cores, and leaves B's cores idle? I noticed the difference between A and B dropped about 3-4ºC when I did that, too. Still I think accessing the core and redoing the thermal paste will be the first order of business once I can get the rivet off that provides the access. I was told that this one overheating CPU didn't used to be a problem, and I wonder if the rigors of shipping could cause ancient thermal paste to crumble and fail?

Strangely, every time it reaches "overtemp" (the light goes on, which usually doesn't crash the system if it's already booted up), I do not feel ANY hot or even warm air coming from the exhausts in the back of the case. That's why I am leery of a possible diode problem, though I suppose it could really be overheating.]

 

[AphoticD]

Thanks, both of you. I've already changed the performance to "reduced". That seems to have dropped the core temps during operation by about 15ºC each, give or take. Under moderate load CPU B is staying in the 60s that way (CPU A is was mostly in the 30s). When the machine has been off for a while and the innards are back to ambient temperature, it will usually stay on and working. But any reboot will fail until the machine has cooled down again.

I actually have another (I think) working quad CPU/LCS core I may try swapping in also. I just want to rule everything else out before I start a teardown; I suspect the calibration may fail, too. But I'll still try it first. I've never attempted it on a G5 so I'm a little gun shy, but it's (largely) a 50-pound brick if I don't. After that I'll do the thermal paste and try again, then I may swap out and see if my other core is better. If so I'll have to overhaul the LCS, probably.

[Edit to add -- got the Processor preference loaded -- I assume switching to "dual core" (using only CPU 1 and CPU 2) corresponds to CPU A's cores, and leaves B's cores idle? I noticed the difference between A and B dropped about 3-4ºC when I did that, too. Still I think accessing the core and redoing the thermal paste will be the first order of business once I can get the rivet off that provides the access. I was told that this one overheating CPU didn't used to be a problem, and I wonder if the rigors of shipping could cause ancient thermal paste to crumble and fail?

Strangely, every time it reaches "overtemp" (the light goes on, which usually doesn't crash the system if it's already booted up), I do not feel ANY hot or even warm air coming from the exhausts in the back of the case. That's why I am leery of a possible diode problem, though I suppose it could really be overheating.]

When it reaches overtemp, it doesn’t necessarily mean the exhaust air is any hotter than usual. It would be a better sign if the heat was being blown out. The trouble is that the heat isn’t being drawn off the core(s) effectively, so the CPU hits 70C+ and causes an overheat failure, while the ambient temps remain consistent. The fan speed might not be ramping up to compensate because the issue is simply that the heatsink(s) aren’t drawing heat, so a faster fan might not even make a difference.

The shipping would definitely have caused aged, powdered, cracked paste to shake out of place.

New paste should be easy enough. It’s no big deal to just cut the plastic rivet off with a blade. I did that on my Dual Core G5 instead of messing around with it every time I want to get at the CPUs

 

[ziggy29]

When it reaches overtemp, it doesn’t necessarily mean the exhaust air is any hotter than usual. It would be a better sign if the heat was being blown out. The trouble is that the heat isn’t being drawn off the core(s) effectively, so the CPU hits 70C+ and causes an overheat failure, while the ambient temps remain consistent. The fan speed might not be ramping up to compensate because the issue is simply that the heatsink(s) aren’t drawing heat, so a faster fan might not even make a difference.

Fans are maxing out. So it sounds as though *likely* culprits are bad paste and/or a bad LCS that needs to be fixed (knowing nothing can be diagnosed remotely like this)? More likely than a bad diode or temperature sensor on the CPU? I'm running it right now to respond here, CPU A is at 42ºC and CPU B is at 81ºC. Don't want to run it for too long at that temp, though. I'm inclined to fix the paste first, then replace the CPU/LCS core with a spare I think is good if that didn't help.

Here's a current shot. In reality I can hear the fans roar and pull back, often, but the sensors here always show them maxed out. This is with "reduced" processor performance in Energy Saver and with CPU3 and CPU4 shut off in the Processors preference!

 

[AphoticD]

Fans are maxing out. So it sounds as though *likely* culprits are bad paste and/or a bad LCS that needs to be fixed (knowing nothing can be diagnosed remotely like this)? More likely than a bad diode or temperature sensor on the CPU? I'm running it right now to respond here, CPU A is at 42ºC and CPU B is at 81ºC. Don't want to run it for too long at that temp, though. I'm inclined to fix the paste first, then replace the CPU/LCS core with a spare I think is good if that didn't help.

Here's a current shot. In reality I can hear the fans roar and pull back, often, but the sensors here always show them maxed out. This is with "reduced" processor performance in Energy Saver and with CPU3 and CPU4 shut off in the Processors preference!

Switch the poor thing off and go grab some thermal paste from your local electronics store. Anything rated over 5w/mk will work to prevent the CPU from frying itself. 80°C+ is likely to damage the CPU and/or logic board.

Once you've got all 4 cores running around the same temp, then invest in overhauling the LCS and/or a more expensive thermal paste.

 

[ziggy29]

Yes, it's off now. Will get the tools needed to remove the processor core and LCS, already have the thermal paste left over from my PC build a few months ago (Noctua NT-H1). Will see about replacing the paste, or maybe swapping another CPU/LCS core I have available.

 

[ziggy29]

Should have the tools tomorrow -- need only two, I believe, a 3 mm standard hex and a 4mm ball hex (both with long handles/shafts to get in there).

As I understand it, there are screws on the underside of the core assembly to separate out the metal plate from the CPU and that is where the thermal paste connection is? I'd unscrew the bottom plate, remove the old thermal paste, add the new paste and reassemble?

 

[AphoticD]

Should have the tools tomorrow -- need only two, I believe, a 3 mm standard hex and a 4mm ball hex (both with long handles/shafts to get in there).

As I understand it, there are screws on the underside of the core assembly to separate out the metal plate from the CPU and that is where the thermal paste connection is? I'd unscrew the bottom plate, remove the old thermal paste, add the new paste and reassemble?

Yes, that sounds about right. I haven't worked on an LCS unit, but if they are anything like the air cooled CPU card designs, then yes it should be a matter of using the same tools to remove the CPU cards from the underside of the LCS once the Quad CPU unit is out of the tower.

 

[ziggy29]

Took the CPU out, was going to remove heatsink -- seems stuck somehow, didn't want to break it. Will investigate later. In the meantime I also removed a CPU core from another G5 Quad for parts, put it in the working unit and... so far, so good!

The replacement unit is a single-pump LCS Delphi unit, where as the one I took out was a dual pump Cooligy/Delphi unit.

[Edit to add: Since I swapped in a new CPU/LCS core, I wanted to run Thermal Calibration off of ASD 2.6.3 and I was able to boot into the diagnostic through Open Firmware, but I get no cursor and have no mouse control. Has anyone else seen that happen before?]