Help with Quad G5 hardware issue

[dorkshoei]

I have a late 2005 Quad Core 2.5Ghz G5.
It has the updated Panasonic cooler.
There are no leaks I can observe.

I'm a Linux developer. I was given the machine by it's original purchaser.
It has always run Linux. I do not have the original Mac disks.
I know close to zero about Macs but I like the Power arch.

-----------------------------

I've not powered in on in a while. I just tried a few days ago and noticed while it would still boot into open-firmware (OF) it no longer chimed at bootup. It seemed stable in open firmware.

I should have left it alone

Instead to try and resolve the lack of chime I tried some of the tips I found online; remove PRAM 3V (cr2032) overnight; boot with option-cmd-p-r; in OF run "reset-nvram", "set-defaults", "reset-all", hold down power button thru programming tone and chime into OF etc etc.

Unfortunately at some point things got a lot worse .... the machine is now totally unstable in OF. It hangs constantly. From a cold boot "dev / ls" will either lockup on first run or second. If I leave it at prompt for >1 minute it will have locked up. Fans come on full shortly after. If I repeatedly run something like "printenv" it will eventually lockup.

"setenv" (such as "setenv auto-boot?") has stopped working, it either hangs after reporting ok or does nothing.

If I was to write a Linux installer to USB there is no way I'd be able to run the necessary OF commands to locate the correct device and boot the installer before it locks up.

I can still boot the installed Linux (from SATA) and oddly if I run "/bin/nvram --update-config auto-boot? false" it actually works (unlike "setenv" in OF) and the machine will auto-boot into OF and printenv shows it is now false.

Under Linux the machine reports all 4 cpus. It seems stable in XWindows. I can ssh in, do various low load tasks without issue in X. However I wrote a short single-threaded C program to factor primes and after about 10 minutes running one instance in a sequential loop the machine abruptly powers itself off without warning (100% reproducible).

I downloaded ASD 2.63. I couldn't figure out how to boot it from OF but from power on if I hold 'C' the all grey screen appears (which I see in various Youtube videos). It then either hangs completely (screen all grey; no picker; fan comes on after 1 minute) or reboots from the SATA and grub appears.

I just tried it now, machine has been off for 12 hours. Held down c. Hear CD spinning up. Grey screen appeared, Then it rebooted and grub appeared. I was able to boot into Linux. I was able to run "mount -t hfsplus /dev/cdrom /mnt" and I can see the ASD 2.63 files. Machine has uptime of 1 hour as I write this.

I need to find a machine that uses DDR2 to test the RAM.

On other forums I've been told I have a coolant leak, the machine is overheating, the cpu's are dying. Help, suggestions, commiserations welcomed.

 

[dorkshoei]

Machine has now been up for >3 hours.

So it seems there is an overheating issue. I'm struggling to however understand how it could be causing open firmware to behave so erratically immediately after a cold boot and causing ASD to hang at cold boot.

# uname -a
Linux powermac 3.0.101-55-ppc64 #1 SMP Mon Apr 20 05:06:00 UTC 2015 (c09be56) ppc64 ppc64 ppc64 GNU/Linux

# uptime
08:19am up 3:10, 5 users, load average: 1.52, 1.35, 1.27

# cd /sys/devices/platform/windfarm.0

# for i in *; do [ -f $i ] && echo "$i: `cat $i`"; done
backside-fan: 1100
backside-temp: 57.000
cpu-current-0: 6.710
cpu-current-1: 6.957
cpu-current-2: 9.152
cpu-current-3: 9.886
cpufreq-clamp: 0
cpu-front-fan-0: 3104
cpu-front-fan-1: 3104
cpu-power-0: 8.241
cpu-power-1: 8.560
cpu-power-2: 12.535
cpu-power-3: 13.179
cpu-pump-0: 3584
cpu-pump-1: 3584
cpu-rear-fan-0: 3200
cpu-rear-fan-1: 3200
cpu-temp-0: 39.109
cpu-temp-1: 39.156
cpu-temp-2: 63.171
cpu-temp-3: 67.531
cpu-voltage-0: 1.228
cpu-voltage-1: 1.230
cpu-voltage-2: 1.369
cpu-voltage-3: 1.333
drive-bay-fan: 1000
hd-temp: 28.500
slots-fan: 1560
slots-power: 18.417

Half an hour ago I started Firefox and in the log I see:
Apr 8 07:57:44 powermac kernel: windfarm: Overtemp due to immediate CPU temperature !
Apr 8 07:57:44 powermac kernel: windfarm: Clamping CPU frequency to minimum !
Apr 8 07:57:45 powermac kernel: windfarm: Clamping CPU frequency to minimum !
Apr 8 07:58:19 powermac kernel: windfarm: Overtemp condition cleared !
Apr 8 07:58:20 powermac kernel: windfarm: CPU frequency unclamped !

As I said above, it does seem to be overheating. I see this (from about the time yesterday when I ran the prime load test and the machine abruptly powered itself off):

Apr 7 17:28:54 powermac kernel: windfarm: Overtemp due to immediate CPU temperature !
Apr 7 17:28:54 powermac kernel: windfarm: Clamping CPU frequency to minimum !
Apr 7 17:28:55 powermac kernel: windfarm: Clamping CPU frequency to minimum !
Apr 7 17:29:50 powermac kernel: windfarm: Critical overtemp due to immediate CPU temperature !

 

[MysticCow]

Linux

There’s your problem!

If it’s overheating, you need to tear it down, reapply thermal paste, and run the Apple Hardware CD you can find on Mac Garden.

That should handle it, but you also got the sucktacilar Liquid Cooled G5s. Those are just plain icky due to liquid leaks and such. There is a way to convert it to air cooled, but it still takes thermal paste and the Apple CD from Mac Garden.

 

[dorkshoei]

There’s your problem!

If it’s overheating, you need to tear it down, reapply thermal paste, and run the Apple Hardware CD you can find on Mac Garden.

A URL would help as if you reread I aalready have a copy of the ASD disc.

That should handle it, but you also got the sucktacilar Liquid Cooled G5s. Those are just plain icky due to liquid leaks and such.

The quad uses the updated Panasonic cooler.

i believe you are thinking of earlier G5 models which used the Delphi cooler which did leak.

I've not found a report of a leaking Panasonic. I'm sure there has been a case but everything I read leads me to believe they are reliable and some people are just thinking of the earlier Delphi coolers

 

[joevt]

I use setenv boot-args cpus=3 in Open Firmware to make macOS use only 3 processors instead of the normal 4 for my Quad G5 otherwise it will hang. It's probably a lucky coincidence that the one CPU that is disabled is the one that causes the hang. Otherwise a more complicated Open Firmware hack would be required.

Is there a similar setting for the Linux kernel?

My Open Firmware is stable while yours is not, so reducing CPUs might not be sufficient.

 

[dorkshoei]

I use setenv boot-args cpus=3 in Open Firmware to make macOS use only 3 processors instead of the normal 4 for my Quad G5 otherwise it will hang. It's probably a lucky coincidence that the one CPU that is disabled is the one that causes the hang. Otherwise a more complicated Open Firmware hack would be required.

Is there a similar setting for the Linux kernel?

My Open Firmware is stable while yours is not, so reducing CPUs might not be sufficient.

First thanks for trying to answer the question and I appreciate the first hand info ... I'm curious; have you tried replacing one of your cpus?

https://www.kernel.org/doc/html/latest/admin-guide/kernel-parameters.html maxcpus= sets the limit on how many can be brought up. You can offline specific cpus at runtime, I'm not sure if you can specify specific cpus to enable on the command line. I'd have to check, plus I'm running a really old kernel.

But yes, unless one can extend this to open firmware, it won't help me.

Clearly there is an overheating issue (as Linux reports it) but I'm very surprised at how unstable open firmware is. I don't see how the Forth interpreter inside open firmware could be a major resource hog but hardware is a mysterious thing.

I could be chasing two issues, but clearly I should first try and fix the known case (Linux kernel reports a critical temp issue under high(er) cpu load).

Thanks!

 

[joevt]

I'm curious; have you tried replacing one of your cpus?

Nope. I don't know how much that would cost and I don't want to touch the liquid cooler. I hardly ever use it. It works as is. Meat Loaf said 2 out of 3 aint bad, and 3 out of 4 is better than that.

 

[Dronecatcher]

A problem I had with one of my G5s was that the CPU thermal sensors were broken (as reported by the ASD disk) so the system thought it was going into meltdown when it wasn't.

My solution though not ideal was to run it in reduced power mode (from System Preferences) so the 2.3Ghz processors were running around 1.6Ghz.

 

[dorkshoei]

Nope. I don't know how much that would cost and I don't want to touch the liquid cooler. I hardly ever use it. It works as is. Meat Loaf said 2 out of 3 aint bad, and 3 out of 4 is better than that.

Yeah. Can't argue with this. No point touching it until it gets worse.

A problem I had with one of my G5s was that the CPU thermal sensors were broken (as reported by the ASD disk) so the system thought it was going into meltdown when it wasn't.

My solution though not ideal was to run it in reduced power mode (from System Preferences) so the 2.3Ghz processors were running around 1.6Ghz.

I wish I could boot the ASD disk. As I said, it hangs for me (right after the screen turns all grey post boot).

I just powered it on after 16 hours off. Held opt-cmd-o-f to go into open firmware. Ran "dev / ls" (lists all devices). First time worked, second time worked. Third time it hung mid output. Fans kicked on full after another minute. Total time powered on in open firmware until hang, under one minute.

Linux still boots and runs for multi-hours as long as I don't really load it (prime factoring).

I'm still trying to find an x86 machine which has DDR2 so I can test the RAM.

If the RAM tests good I'm not sure what to do. The open firmware instability to me seems so extreme that it points to some other hardware issue rather than cooling (obviously there is a cooling issue also).

Maybe it is all cooling but without knowing the open firmware and cooling are linked I'm reluctant to put in the time to rebuild the cooling only to have the open firmware issues persist and be forced to junk the machine.

 

[Dronecatcher]

Linux still boots and runs for multi-hours as long as I don't really load it

If you set CPUFreq low I wonder if it'd keep running with no hitches?

 

[dorkshoei]

If you set CPUFreq low I wonder if it'd keep running with no hitches?

Not sure how to do that as I don't have MacOS. Can you do it in OF?

Of course as I mentioned in the first post, "setenv" isn't working any more in OF. It either hangs or does nothing but I can run "nvram" from Linux which does the same and it works. Very odd.

 

[Dronecatcher]

Not sure how to do that as I don't have MacOS. Can you do it in OF?

It's a parameter in Linux - can't remember the exact Terminal syntax - been a while

 

[dorkshoei]

It's a parameter in Linux - can't remember the exact Terminal syntax - been a while

I was more curious if it was possible to throttle the CPUs while in open firmware. To see if the hangs in OF are a result of temperature. To me it's hanging way too soon to be temp related but hardware is funny stuff.

Yes I can likely throttle the CPUs for Linux on kernel cmdline but the issue is that without a working OF it's fairly hard to update to a newer version of Linux, so the machine is pretty useless.

I've not been able to find any other example of someone having hangs like I am in open firmware or having the ASD disk lockup.

 

[Dronecatcher]

I've not been able to find any other example of someone having hangs like I am in open firmware or having the ASD disk lockup.

Have you tried it with the HDD disconnected?

I know that doesn't really make sense but I've had two Macs that had bad drives that prevented booting from any other media or port - usb and firewire drives were simply invisible!

 

[dorkshoei]

Have you tried it with the HDD disconnected?

I know that doesn't really make sense but I've had two Macs that had bad drives that prevented booting from any other media or port - usb and firewire drives were simply invisible!

With hard disk connected, boots into grub (Linux bootloader)

With hard disk connected, holding C key, screen displays all light grey background, hangs, fan comes on full after 1 min (indicating lockup)

With hard disk disconnected, screen displays all light grey background, hangs, fan comes on full after 1 min (indicating lockup)

With hard disk disconnected, holding C key I can hear the CD make a chirp and the drive spins up, then one of four things happens:

- screen displays all light grey background, hangs, fan comes on full after 1 min (indicating lockup)
- screen displays all light grey background, hangs, blank folder symbol appears, fan comes on full after 1 min (indicating lockup)
- screen displays all light grey background, hangs, folder symbol appears with ? symbol inside it, fan comes on full after 1 min (indicating lockup)
- screen displays all light grey background, hangs, folder symbol appears with rotating symbol inside it, first ? then "finder face", then blank, then pattern repeats, after short time pattern ends [lockup probably], fan comes on full after 1 min (indicating lockup)

I suspect these 4 are all the same just a matter of how fast it locks up and so how much of the rotating sequence is displayed. DVD drive spins down fairly quickly, likely as soon as it actually hangs. Followed 30-40 seconds later by the fans coming on full.

I can find examples online of the folder with the ? inside but nothing which discusses a rotating ? and folder face. I suspect it doesn't like the DVD? I converted the dmg to an iso using dmg2img (http://vu1tur.eu.org/tools/). I can mount it fine from Linux using mount -t hfsplus.

Of course the fact that it hangs so fast while displaying the folder image doesn't bode well for actually booting the DVD if I fix whatever it doesn't like.

I just found a PC that takes DDR2 so hopefully I can test the RAM. Removing one half of the RAM didn't fix any of the open firmware issues neither did taking out the remaining two sticks and swapping in the first pair removed so I'm not optimistic it's bad RAM.