Problems with kernel panics when sleeping since 10.6.8

The new memory upgrade did not go smoothly. Upon inserting all 8 DIMMs I got a red fault LED on one of them along with a blinking power LED and no boot. The fault followed this particular DIMM, so I will send it back for replacement. Luckily I have a pair of old 2GB DIMMs to replace it with. So now I have 16GB and the machine boots fine.

I ran the Apple Hardware Test twice (full test) on the sytem and no errors were reported. I read on-line that AHT isn't exactly meticulous in it's inspection, so I downloaded Memtest-86 to try it out. I wasted hours trying to figure out what the heck was going on. NONE of my memory, old or new, would pass the tests. I yanked everything out and ran the test on a pair of my original 512MB Micron DIMMS. When those failed, I threw the Memtest-86 CD in the trash. Others have reported issues with that software, but I was hoping v4.0a had all that fixed. I guess not. If there really is a memory issue, I would now be more inclined to point at my memory controller than 16 bad DIMMs. I hope that is not the case and that software is just buggy. AHT was happy, so I put all of the new mem back in (except for the 1 DIMM I replaced) and now I will give it two weeks to see if it can stay up without a kernel panic or lockup.

Memtest is part of Rember - puts a GUI around it.

ECC is hard to test.

Memtest can be run multiple concurrent in Terminal, with each doing a smaller chunk letting you run 32 instances

Lion Cache Cleaner (was sLCC, backward compatible) also includes Memtest

There are multple versions but sounds like you got the wrong one. Windows 7 and Vista also have memtest on their DVD.

Thanks. I did see that other Memtest on the web, but it seemed more complicated (run in single user mode) and the author didnt even sound very confident in it (shave off about 2.8% of your actual memory to prevent crashing). So I went with the one that booted from CD into a small linux distro. Not sure what makes memory so hard to test. Once I get the replacement DIMM I will try again with something else.

Rember seems to

Take longer because it is GUI and single instance;

Reliable

Single User Mode is so you can test more without OS and kernel having to load (not sure but 80MB?) while I would just swap DIMMs in/out and around.

Memtest in multiple terminal windows though is what I do, testing 333MB chunk for 5 loops, and enough windows to test everything.

Memtest OS X

http://www.macports.org/ports.php?by=name&substr=Memtester

Rember

Comments etc http://www.macupdate.com/app/mac/15837/rember

Lion/Leopard/Panther/Tiger Cache Cleaner bundled and kept current plus other features. And has its own Single User Mode (when AppleJack support for Intel was missing).

But.. Memtestx86 never really worked for Intel Macs and esp. for ECC memory - I guess tricking the RAM to fail? is hard.

Bottom line: Rember and watch System Profile: Memory.

AHT does spot mechanical hardware errors with RAM and memory daughter/Riser cards, so if there was a parity error, AHT would spot those. Or any sensors. FBDIMMs were critcial when it came to temperatures.

Rember scan (3 times) on Sunday indicated no issues with the new memory. The system locked up on Wednesday, so obviously the PRAM and SMC reset had no effect. The CPU temp on the frozen display was 92 deg F, so I doubt that was an issue. The only thing I can think to try is to yank Parallels.

Parallels uninstalled using the v6 uninstaller and the 'rm' commands listed on the Parallels website KB article for uninstall (kext removal and such). Rebooted and installed folding@home to provide some system stress. So far, up for 10 hours and no issues. We'll see if it lasts a week. Not sure what to do if it does pass. Switch to VMware? Upgade to Parallels v7?

So without Parallales 6 installed it crashed thrice this week. Once on Wednesday, and twice again on Thursday:

Wed Sep 21 06:43:40 2011

panic(cpu 1 caller 0x2b0571): "TLB invalidation IPI timeout: " "CPU(s) failed to respond to interrupts, unresponsive CPU bitmap: 0x10, NMIPI acks: orig: 0x0, now: 0x0"@/SourceCache/xnu/xnu-1699.22.81/osfmk/i386/pmap.c:3078

Backtrace (CPU 1), Frame : Return Address (4 potential args on stack)

Thu Sep 22 07:47:08 2011

panic(cpu 5 caller 0x2c3692): "Spinlock acquisition timed out: lock=0x863148, lock owner thread=0x17b3dbe8, current_thread: 0x195277f0, lock owner active on CPU 0x0"@/SourceCache/xnu/xnu-1699.22.81/osfmk/i386/locks_i386.c:375

Backtrace (CPU 5), Frame : Return Address (4 potential args on stack)

Thu Sep 22 20:30:19 2011

panic(cpu 7 caller 0x2b0571): "TLB invalidation IPI timeout: " "CPU(s) failed to respond to interrupts, unresponsive CPU bitmap: 0x30, NMIPI acks: orig: 0x0, now: 0x0"@/SourceCache/xnu/xnu-1699.22.81/osfmk/i386/pmap.c:3078

Backtrace (CPU 7), Frame : Return Address (4 potential args on stack)

My only hope now is to roll-back to OS X v10.6.7 or perhaps put my dual core CPU's back in. This s-u-c-k-s. I think I will do the CPU swap since I would have more use of Lion than I would the 4 extra cores. When I do run multiple VM's, they are lightly loaded.

From one of the links above I stumbled into this thread on Netkas:

http://forum.netkas.org/index.php/topic,1094.0.html

I successfully updated my MacPro 1,1 system to the MP21.007F.B06 firmware that is for MacPro 2,1 systems. Since those are legitimate 8-core systems, the logic is that there is updated microcode in the firmware that better supports a quade core CPU. The response has been mostly positive.

I upgraded my system tonight and re-installed Parallels. I'll see if it can run for a week without crashing or locking up on me.

I really hope this fixes it.

I've had 3 KP's since I upgraded my 2006 Pro with a pair of X5365 procs. Last two while printing.

I found the link you mentioned a coupld of days ago, and installed it, not my machine thinks it's a 2007 model (Mac 2,1). So far no KP's but I'll let you know too if mine are gone as well.

The firmware fix does make a whole lot of sense to me.....

Something happened Wednesday morning. My uptime was reset, but there was no KP log and no notice that the computer had reset. It just rebooted. I supposed it's possible there was a power outtage, but I think that is unlikely. So, this firmware didnt fix the issue and now I can either downgrade my CPU's or OS. Lame.

I also upgraded to Parallels v7 and it was running when the reboot happened, so don't expect that to fix anything either. 😟

I just wanted to follow up and say that since that one reboot (which was probably power related since there was no log or KP data) that I have not had any lockups or crashes since upgrading to the 2,1 firmware. My server has been up nearly 9 days and running Parallels 7 and OS X Lion the entire time.

I think this is a legitimate fix for any Mac Pro 1,1 that has had the CPU's upgraded to quad core and wants to run OS X 10.6.8 or greater.

I've been up for around 2 weeks now since the 2,1 firmware, and not a single issue either. Did the 10.6.8 supplemental update, and it's been the only time I've had to reboot the machine. And I've been giving Handbrake quite the workout since the update. Nice to see 8 cores running full blast! A 2 hour movie take less than 15 mintes to rip now.

For now I am staying away from Lion, I just don't have a need for it right now. When they get to 10.7.2 or 3 I may consider it.

Sleeping with 5870 and multiple monitors causes my issues.

This is definately the problem with my machine, where I have three monitors attached. I was manually turning off two of the monitors and then letting the machine go to sleep itself.

The situation was definately better with Snow Leopard, it has become worse with Mountain Lion (I never wnet through Lion on this machine).

I have all three monitors attached to the one card, but will be testing a second card soon to run the other screens.

Do you have a panic log report?

The machine just does not wake up. No curtain type kernel panic, or kernel reports. Just lots (LOTS!!!) of entries in the log

ATICLASS::detectPowerDown: GPU is not found. PCI config access fails. VendorID = 0xFFFF !!!

The machine will go to sleep, it will just never wake up, with no network or remote access either.

This is on a MacPro3,1 Early 2008 runnin 10.8.3, Radeon 5870, Three screens, two via Mini Display Port to Dual-Link DVI adapters (the big USB powered ones)

Message was edited by: Charlie Nancarrow