Resolved Faulty CPU? X5690 in MP5,1

[phybron]

I've just swapped my X5670's for X5690's. The machine boots and works, but if I try to run Geekbench it crashes, shuts down and restarts every time. I managed to run Novabench, so that's not stressing the system in the same way.

I have upgraded the CPU's in this machine three times in the past and have never had this problem. Is it indicative of a faulty CPU? This is the first time I have not specifically bought a 'matched pair' as I'm unsure that means anything.

I tried to run Apple Hardware Test and that crashed it as well.

Crash report looks like this:

*** Panic Report ***
Panic(CPU 2): NMIPI for spinlock acquisition timeout, spinlock: 0xffffff80399f9948, spinlock owner: 0xffffff803b5cb6d0, current_thread: 0xffffff803b5cb6d0, spinlock_owner_cpu: 0x2
RAX: 0x000000004a1485f9, RBX: 0xffffff8037fc2800, RCX: 0x0000000000000003, RDX: 0x000001584a1485f9
RSP: 0xffffff92d4d0bbe0, RBP: 0xffffff92d4d0bc10, RSI: 0x00000177e305b1ab, RDI: 0xffffff82c7ef6148
R8:  0xffffff800bcddb78, R9:  0xffffff800bcddb78, R10: 0x0000000000000000, R11: 0x0000000000000246
R12: 0x0000000000000000, R13: 0xffffff8037fc8400, R14: 0xffffff82c7ef6148, R15: 0x0000000000000001
RFL: 0x0000000000000002, RIP: 0xffffff7f8de9db06, CS:  0x0000000000000008, SS:  0x0000000000000010
Backtrace (CPU 2), Frame : Return Address
0xffffff82c7ec6f80 : 0xffffff800b5d5e0f 
0xffffff82c7ec6fd0 : 0xffffff800b5f4259 
0xffffff92d4d0bc10 : 0xffffff800b5dfa38 
0xffffff92d4d0bc30 : 0xffffff800b51ec2f 
0xffffff92d4d0bcf0 : 0xffffff800b520dd0 
0xffffff92d4d0bd30 : 0xffffff800b9900e5 
0xffffff92d4d0bdc0 : 0xffffff800b99215f 
0xffffff92d4d0bf00 : 0xffffff800b98f7b2 
0xffffff92d4d0bf60 : 0xffffff800ba2ace1 
0xffffff92d4d0bfb0 : 0xffffff800b5f4896 
      Kernel Extensions in backtrace:
         com.apple.driver.AppleIntelCPUPowerManagement(218.0)[5694EADA-D8D4-3FFA-86BD-172FDCC65526]@0xffffff7f8de9a000->0xffffff7f8dec4fff

BSD process name corresponding to current thread: mds

Mac OS version:
15A284

Kernel version:
Darwin Kernel Version 15.0.0: Wed Aug 26 16:57:32 PDT 2015; root:xnu-3247.1.106~1/RELEASE_X86_64
Kernel UUID: 37BC582F-8BF4-3F65-AFBB-ECF792060C68
Kernel slide:     0x000000000b200000
Kernel text base: 0xffffff800b400000
__HIB  text base: 0xffffff800b300000
System model name: MacPro5,1 (Mac-F221BEC8)

System uptime in nanoseconds: 398539658139
last loaded kext at 244737776893: com.apple.filesystems.msdosfs    1.10 (addr 0xffffff7f8e2e6000, size 69632)
last unloaded kext at 305392309650: com.apple.filesystems.msdosfs    1.10 (addr 0xffffff7f8e2e6000, size 61440)
loaded kexts:
com.nvidia.CUDA    1.1.0
com.apple.driver.AppleHWSensor    1.9.5d0
com.apple.driver.AppleTyMCEDriver    1.0.2d2
com.apple.driver.AGPM    110.20.19
com.apple.filesystems.ntfs    3.13
com.apple.driver.AppleOSXWatchdog    1
com.apple.filesystems.autofs    3.0
com.apple.driver.AppleUpstreamUserClient    3.6.1
com.apple.driver.AppleMCCSControl    1.2.13
com.apple.driver.AppleMikeyHIDDriver    124
com.apple.driver.AppleHDA    272.50
com.apple.GeForce    10.0.6
com.apple.driver.AppleMikeyDriver    272.50
com.apple.driver.pmtelemetry    1
com.apple.iokit.IOUserEthernet    1.0.1
com.apple.iokit.IOBluetoothSerialManager    4.4.0f4
com.apple.Dont_Steal_Mac_OS_X    7.0.0
com.apple.iokit.BroadcomBluetoothHostControllerUSBTransport    4.4.0f4
com.apple.driver.AppleHV    1
com.apple.driver.AppleLPC    3.1
com.apple.driver.AudioAUUC    1.70
com.apple.driver.AppleIntelSlowAdaptiveClocking    4.0.0
com.apple.driver.ACPI_SMC_PlatformPlugin    1.0.0
com.apple.iokit.IOAHCIBlockStorage    2.8.0
com.apple.iokit.SCSITaskUserClient    3.7.7
com.apple.AppleFSCompression.AppleFSCompressionTypeDataless    1.0.0d1
com.apple.AppleFSCompression.AppleFSCompressionTypeZlib    1.0.0
com.apple.BootCache    37
com.apple.driver.AppleFWOHCI    5.5.2
com.apple.driver.

 

[SDAVE]

May be a matched pair issue.
May be a bad thermal paste job.
May be faulty memory.

 

[phybron]

May be a matched pair issue.
May be a bad thermal paste job.
May be faulty memory.

So matched pairs are a thing? It actually makes a difference? I have taken the CPU out and checked the thermal paste. I have the requisite thin layer covering the entire surface. My CPU A is running a bit warm: 117°F. CPU B is at 97°F with Ambient at 88°F.

EDIT: I don't think it's memory because it's the same memory I was using with the old CPU's and had no problems. Hard to tell when I can't run AHT.

 

[SDAVE]

Don't cover the entire surface.

Clean your surface with 99% rubbing alcohol. Dry. Put a pea sized dab in the middle of the CPU(s) and then gently apply the fan (evenly) one turn per corner screw at a time until it is tight (don't tighten too much).

Not 100% sure about the matched pair being an issue, but if it restarts during CPU utilization, it may be an overheating related issue.

Try taking out the memory modules and just running bare with minimum memory possible (it's dual channel, so 1 memory module per CPU) and run the test again after applying CPU compound again.

 

[phybron]

Don't cover the entire surface.

Clean your surface with 99% rubbing alcohol. Dry. Put a pea sized dab in the middle of the CPU(s) and then gently apply the fan (evenly) one turn per corner screw at a time until it is tight (don't tighten too much).

Not 100% sure about the matched pair being an issue, but if it restarts during CPU utilization, it may be an overheating related issue.

Try taking out the memory modules and just running bare with minimum memory possible (it's dual channel, so 1 memory module per CPU) and run the test again after applying CPU compound again.

Ugh, I did the small amount the first time and then added a bit more when I saw it hadn't covered the entire surface. I'll try removing memory first and then clean and redo the CPU. I'm glad this is the last time I'll have to do this, X5690 being the final option for these, as far as I can tell.

Thanks!

 

[SDAVE]

Here are instructions on how to apply thermal paste, straight from the Mac Pro 5,1 service manual used by Apple.

 

[phybron]

Here are instructions on how to apply thermal paste, straight from the Mac Pro 5,1 service manual used by Apple.

That reminds me, I actually have a copy of the service manual. Maybe I should read it again...

In the meantime I pulled RAM and retested Geekbench. It worked with one DIMM for each processor, then two, then three, which is all I have. So, in other words, it seems to work now. I'm a bit concerned about stability, but I'll see how it goes I suppose. I should probably run AHT as well.

 

[SDAVE]

That reminds me, I actually have a copy of the service manual. Maybe I should read it again...

In the meantime I pulled RAM and retested Geekbench. It worked with one DIMM for each processor, then two, then three, which is all I have. So, in other words, it seems to work now. I'm a bit concerned about stability, but I'll see how it goes I suppose. I should probably run AHT as well.

Max out the CPUs at 100% and leave it on overnight. Make sure there's good ventilation around the tower. Come back in the morning and see if it has crashed, if not then you're good.

 

[phybron]

Max out the CPUs at 100% and leave it on overnight. Make sure there's good ventilation around the tower. Come back in the morning and see if it has crashed, if not then you're good.

That's a pretty heavy test. I just ran a few iterations of Geekbench Stress Test and it didn't crash. So maybe it was just a RAM seating issue. CPU A is still running pretty warm, but that could be for all sorts of reasons. I know Yosemite changed the fan ramp-ups from Mavericks in a way that caused everything to run hotter for longer. Maybe El Capitan has done something similar. The fans didn't speed up much during the stress tests.

Thanks for the help.

 

[SDAVE]

That's a pretty heavy test. I just ran a few iterations of Geekbench Stress Test and it didn't crash. So maybe it was just a RAM seating issue. CPU A is still running pretty warm, but that could be for all sorts of reasons. I know Yosemite changed the fan ramp-ups from Mavericks in a way that caused everything to run hotter for longer. Maybe El Capitan has done something similar. The fans didn't speed up much during the stress tests.

Thanks for the help.

You can run MacFanControl and up the fan temps a tiny bit on login. I think the fans hit at around 80c. You can run MacFanControl and have it run at a steady speed so you don't have to worry about throttling. Also remember these CPUs have SpeedStep technology so if things are too hot the CPU will throttle down as well, but usually at around 80c the fans speed up and take care of any heat.

Remember that the 130~watt CPUs were not designed for the Mac Pro, that's why Apple topped at the x5675, so you can bump the fan speed a bit.

But I remember when I had x5680s I never bothered after reseating the CPUs, it ran stable and fans would kick in when needed. Not too loud.