MP 1,1-5,1 RAM(?) issues on MP 4,1->5,1 leading to KPs

[justinr]

I recently got a stock 6-core 2.93 GHz MP 4,1. It seemed to operate normally, and the person I got it from said they never had an issue with it but hadn't used it recently. I wanted to flash it to 5,1 and upgrade to High Sierra in anticipation of upgrading to a 6-core CPU. I followed the procedure listed here and it seemed to work without issue. When I went to install High Sierra, the installer performed an additional firmware upgrade that again proceeded without error. However, under High Sierra I've been experiencing random crashes that are frequently (but not always) attributed to the CPU Power Management kext. I've successfully run 8+ hours of Prime95 and Rember without any errors, high temperatures, or crashes, but something as simple as opening a small jpg in Preview might crash the machine.

Thinking it might be a bad RAM stick, I tried to isolate the problem by removing a stick of RAM and using only slots 1-3. The computer would not boot without the 4th stick, giving no video output and just spinning all the fans on max speed. I tried again with 2 sticks (in slots 1-2) and again it would not boot. Only when all 4 sticks are present will it boot.

I'm probably ordering some replacement RAM tonight, but does anyone have ideas for what else might be going on here? Bad firmware update? Bad CPU tray? Bad CPU? AHT for the 4,1 doesn't tell me anything besides Video Card error (which apparently happens with 4,1->5,1 upgrades).

 

[tsialex]

I recently got a stock 6-core 2.93 GHz MP 4,1. It seemed to operate normally, and the person I got it from said they never had an issue with it but hadn't used it recently. I wanted to flash it to 5,1 and upgrade to High Sierra in anticipation of upgrading to a 6-core CPU. I followed the procedure listed here and it seemed to work without issue. When I went to install High Sierra, the installer performed an additional firmware upgrade that again proceeded without error. However, under High Sierra I've been experiencing random crashes that are frequently (but not always) attributed to the CPU Power Management kext. I've successfully run 8+ hours of Prime95 and Rember without any errors, high temperatures, or crashes, but something as simple as opening a small jpg in Preview might crash the machine.

Thinking it might be a bad RAM stick, I tried to isolate the problem by removing a stick of RAM and using only slots 1-3. The computer would not boot without the 4th stick, giving no video output and just spinning all the fans on max speed. I tried again with 2 sticks (in slots 1-2) and again it would not boot. Only when all 4 sticks are present will it boot.

I'm probably ordering some replacement RAM tonight, but does anyone have ideas for what else might be going on here? Bad firmware update? Bad CPU tray? Bad CPU? AHT for the 4,1 doesn't tell me anything besides Video Card error (which apparently happens with 4,1->5,1 upgrades).

If you cross-flashed to a MP5,1 firmware, you have to use MP5,1 AHT from now on. Anyway, did you started your diagnostics doing the 3-time NVRAM clear? Press CMD-Option-P-R continuously until you hear the fourth chime.

 

[justinr]

Thanks! You seem like the expert on these issues so I really appreciate the reply.

I've tried holding CMD-Option-P-R continuously but the most chimes I ever hear before the machine boots is 2. I have an A1314 bluetooth keyboard, do I need a wired keyboard to make that work? I've tried letting go of the keys and re-pressing them after the chime but it still just boots after the second chime. Also tried the PC keyboard I have on hand but couldn't get it to do it even once (pressing Win-Alt-P-R). Would a MacAlly wired USB keyboard work for this? Apple doesn't sell wired keyboards anymore and that's what I can get locally. Otherwise I'll eBay or CL and older Apple USB wired keyboard. (I do have an ADB to USB keyboard adapter on the way for my AEK II, maybe that will work?)

I did run the 5,1 AHT anyway and it reports no errors.

Any thoughts on why the machine won't boot with anything but 4 sticks of RAM? I can't find anyone reporting a similar kind of error and it makes me think there's some serious hardware or firmware issue. I would try updating the firmware further (currently MP51.0089.B00) using the Mojave installer, but the only Metal-capable card I have on hand (a PowerColor RX590) requires more power than the mini 6 pin connectors can provide.

 

[tsialex]

Thanks! You seem like the expert on these issues so I really appreciate the reply.

I've tried holding CMD-Option-P-R continuously but the most chimes I ever hear before the machine boots is 2. I have an A1314 bluetooth keyboard, do I need a wired keyboard to make that work? I've tried letting go of the keys and re-pressing them after the chime but it still just boots after the second chime. Also tried the PC keyboard I have on hand but couldn't get it to do it even once (pressing Win-Alt-P-R). Would a MacAlly wired USB keyboard work for this? Apple doesn't sell wired keyboards anymore and that's what I can get locally. Otherwise I'll eBay or CL and older Apple USB wired keyboard. (I do have an ADB to USB keyboard adapter on the way for my AEK II, maybe that will work?)

I did run the 5,1 AHT anyway and it reports no errors.

Any thoughts on why the machine won't boot with anything but 4 sticks of RAM? I can't find anyone reporting a similar kind of error and it makes me think there's some serious hardware or firmware issue. I would try updating the firmware further (currently MP51.0089.B00) using the Mojave installer, but the only Metal-capable card I have on hand (a PowerColor RX590) requires more power than the mini 6 pin connectors can provide.

Wireless KBs will not work for multiple clear-NVRAM, needs to be a wired KB.

Any decent wired USB KB will work, you are correct, with a Windows KB the Option key is ALT, while Command is the Windows key. Try to do the 3-times consecutively clear NVRAM before anything. My backup KB is a Microsoft Wired 600, cheap and works every time.

Don't try to update the firmware with a problematic Mac Pro, you can get a brick doing it.

 

[justinr]

I got a wired USB keyboard and was able to do a multiple-clear of the NVRAM. I ran an extended 5,1 AHT and it reported no errors. I used the computer some more and it still crashed. It seems like I can pretty consistently get it to crash by running the Heaven benchmark, but it will crash under other circumstances (like after typing a password to login, or opening a picture file, or other actions) as well. It does seem like it has to be doing something though. It doesn't just crash when it is sitting there with nothing else going on.

I'm considering installing Windows or another OS and seeing if it crashes under that OS as well. I think this has to be a hardware error since it's happened after reinstalling multiple times on multiple SSDs, but it would be nice to rule out completely.

 

[tsialex]

I got a wired USB keyboard and was able to do a multiple-clear of the NVRAM. I ran an extended 5,1 AHT and it reported no errors. I used the computer some more and it still crashed. It seems like I can pretty consistently get it to crash by running the Heaven benchmark, but it will crash under other circumstances (like after typing a password to login, or opening a picture file, or other actions) as well. It does seem like it has to be doing something though. It doesn't just crash when it is sitting there with nothing else going on.

I'm considering installing Windows or another OS and seeing if it crashes under that OS as well. I think this has to be a hardware error since it's happened after reinstalling multiple times on multiple SSDs, but it would be nice to rule out completely.

The best course of action is to methodically test your Mac Pro components with a known working Mac Pro, testing each part. Don't forget to check the Northbridge heatsink clips.

Another thing to try is to run ASD, but it's an internal Apple tool and it's very difficult to locate a copy on the net. MP4,1 ASD is 3S130 and MP5,1 ASD is 3S149, but you need only the latter.

Remember that AHT and ASD don't really test GPUs and the tests done to it are meaningful only for Apple OEM GPUs, so, most times a problematic GPU will not be detected at all.

 

[justinr]

If only I had another Mac Pro to test with! I did get new RAM installed today and so far the system hasn't crashed. The Heaven benchmark used to consistently crash the system within 15 minutes but so far I've ran it for 12 hours without a problem. I'll try a few other things tomorrow to see if it crashes with that, but I think the issue may have just been bad RAM that evaded detection by rember or AHT.

 

[Macschrauber]

Show temperatures with a tool like macsfancontrol.

if Heaven crahses it could be the gpu itself, double check the gpu.

Northbridge cooling could be an issue.

imho its no good idea to stress kill components like running benchmarks forever

 

[justinr]

I was monitoring temps using Hardware Monitor and there was no issue. The problem persisted across different GPUs (and different slots for the GPU). I'll look further into the Northbridge cooling, but again changing the RAM seems to have addressed it.

As for running benchmarks forever, sure, it's not a good idea to just do it for fun. But when you're trying to find intermittent hardware errors stressing the system is sometimes necessary to bring them out. Although if you have suggestions as to how one can trigger such errors in other ways, I'm interested to hear them.