MP 1,1-5,1 cMP 4,1 (5,1) crashing after CPU upgrade

[Macschrauber]

I also see a little too much of the black seal, it should be absolutely flat to ensure the spacers are not too thick.

i second that there is too much thermal paste. Just a thin layer is enough.

 

[sgentile92]

I also see a little too much of the black seal, it should be absolutely flat to ensure the spacers are not too thick.

i second that there is too much thermal paste. Just a thin layer is enough.

What’s the best way to remove that black residue? I’ve read that it’s beneficial to spread the thermal paste before seating the heat sink to ensure complete coverage. Is that step not necessary?

 

[Macschrauber]

take time and a razor blade and scratch the black residue off completely. Clean the remains with isopropyl.

Its not clearly visible but I think on one of your pictures I see a gap between the processor surface and the plastic spacer.

This gap makes the construction too high and the heat sink cannot contact the processor die completely and evenly.

Remove the old paste completely from die and heatsink, clean it with isopropyl, let it dry and attach a thin layer of paste. I take the PCI bracket as a nice tool for it, you can hold it by the screw and it is handy.

 

[tsialex]

You have to remove every residue of the black sealant or your heatsink won't be flat when using the black plastic spacers.

 

[sgentile92]

Update:
I ran a bunch of tests this weekend. Reinstalled original graphics card, pulled PCI cards, pulled my SSD and 4TB HD, and clean installed 10.13.6 onto the original HD that came with the machine. I tested all kinds of RAM configurations and this is what I have figured out:

Two sticks of my OWC 4GB 1333 RAM cause no crashing when running the Terminal "yes" stress test of all 24 cores (12 physical, 12 virtual, activity monitor reports 99–100% CPU usage) and GB5 benchmark test. I kept the same stick in slot 1 and alternated all 3 others through slot 5. No crashes. As soon as I add the remaining two OWC sticks into slots 2 and 6, GB5's benchmark test causes a crash, BUT there's still no crash during the Terminal stress test. Here's where it gets more interesting: I found two of the original 1GB RAM sticks and there's no crash during either GB5 or the Terminal test. I had been wondering if the 2 and 6 RAM slots were defective but this test would seem to indicate that's not the case. For the hell of it I threw my two remaining 4GB sticks into slots 3 and 7, for a total of 34 GB of RAM and GB5 still triggers a crash. Does it seem strange that individually when I test the 4GB RAM sticks #1 + #2., #1 + #3. #1 + #4, they all work, but when used together the system crashes? I can't figure out what that means.

I then reinstalled the CPUs, carefully scraping off as much of the gunked-up black residue as possible and applying as thin of a coat of thermal paste as I could. Still crashing the same as before. See photos below.

I'm very confused on why GB5's benchmark test crashes the system but the Terminal "yes" command does not. I tested GB4—same result. Prior to removing my SSD and 4TB drive, I ran a baseball simulation using a game called Out of the Park Baseball to simulate years of baseball games and statistics the computer didn't crash/restart, it just froze after a 15 or so minutes of processing that data. I'm seriously dumbfounded at this point. None of the hardware I pulled out to get down to a bare bones system caused anything to change so I'm fairly confident that those elements are not the problem. Is there some other troubleshooting methodology that would better serve me?

Spoiler: Photos

 

[amstel78]

I wonder if there's a problem with your 4,1/5,1 setting RAM timing and or voltages with more than 2 sticks at play?

Perhaps the terminal test isn't stressing other components enough for the issue to manifest whereas GB is?

Can you run Memtest86 with all four sticks? This will let you test just the RAM modules outside of MacOS.

MemTest86 - Download now!

Download your free copy of MemTest86 here for USB.

www.memtest86.com

 

[sgentile92]

I wonder if there's a problem with your 4,1/5,1 setting RAM timing and or voltages with more than 2 sticks at play?

Perhaps the terminal test isn't stressing other components enough for the issue to manifest whereas GB is?

Can you run Memtest86 with all four sticks? This will let you test just the RAM modules outside of MacOS.

MemTest86 - Download now!

Download your free copy of MemTest86 here for USB.

www.memtest86.com

Memtest86 has been running for about 2 hours and on test 13, pass 1/4 it did spit out some errors. What does this tell me exactly? I found this article about getting errors on test 13 (specifically the "Why do I get errors only when testing RAM modules together, and not when individually tested?" section) but I'm not really sure what it means.

PassMark MemTest86 - Memory Diagnostic Tool - Troublingshooting Memory Errors

MemTest86 is a memory test for x86 computers.

www.memtest86.com

Should I let the test keep running?

 

[tsialex]

Memtest86 has been running for about 2 hours and on test 13, pass 1/4 it did spit out some errors. What does this tell me exactly? I found this article about getting errors on test 13 (specifically the "Why do I get errors only when testing RAM modules together, and not when individually tested?" section) but I'm not really sure what it means.

PassMark MemTest86 - Memory Diagnostic Tool - Troublingshooting Memory Errors

MemTest86 is a memory test for x86 computers.

www.memtest86.com

Should I let the test keep running?
View attachment 1737255

Defective memory, it's triggering ECC. I'd replace it right now, it's cheap enough to not think twice, I'd lose a lot more money with a long job that fails.

 

[sgentile92]

Defective memory, it's triggering ECC. I'd replace it right now, it's cheap enough to not think twice, I'd lose a lot more money with a long job that fails.

I'd be over the moon if this finally the solution to my issues. I've been on and off trying to solve it for going on 3 years now. This is my second set of RAM from OWC and I've recently realized they they don't seem to have great reputation and that quite a few people report defective products from them. I've seen this site recommended in a few places:

Mac Pro Memory - 2012 6-Core 3.33GHz MD770LL/A CTO - DMS

DMS offers DDR3-1333 (PC3-10600) Memory for your 6-Core 3.33GHz 2012 Mac Pro MD770LL/A CTO. 100% Apple Compatible and backed by a Lifetime Warranty.

www.datamemorysystems.com

Is there someplace else I should be looking into as a RAM supplier? Looks like I can get (2) 3 x 8GB sets for $240 from them.

 

[tsialex]

I'd be over the moon if this finally the solution to my issues. I've been on and off trying to solve it for going on 3 years now. This is my second set of RAM from OWC and I've recently realized they they don't seem to have great reputation and that quite a few people report defective products from them. I've seen this site recommended in a few places:

Mac Pro Memory - 2012 6-Core 3.33GHz MD770LL/A CTO - DMS

DMS offers DDR3-1333 (PC3-10600) Memory for your 6-Core 3.33GHz 2012 Mac Pro MD770LL/A CTO. 100% Apple Compatible and backed by a Lifetime Warranty.

www.datamemorysystems.com

Is there someplace else I should be looking into as a RAM supplier? Looks like I can get (2) 3 x 8GB sets for $240 from them.

Completely crazy price. Buy server pulls from eBay. You can get 48GB (3x16GB) of the best Samsung RAM for half this price.

Compatible 3rd party RAM:
Samsung M393B2G70DB0-YK0 16GB PC3-12800R 1600MHz 1.35V modules with ECC that operate at 1333MHz in MacPro5,1 with 2x5690 CPUs. Cisco server pulls (Cisco Part Number 15-14595-01). Some other identification stats/specs, if helpful: 16GB DDR3-1600 RDIMM PC3L-12800R Dual Rank x4 Module, CL11, 2Rx4, 240pin

M393B2G70DB0-YK0 SAMSUNG 16GB 2RX4 PC3L-12800R 1.35V MEMORY MODULE (1x16GB) | eBay

Find many great new & used options and get the best deals for M393B2G70DB0-YK0 SAMSUNG 16GB 2RX4 PC3L-12800R 1.35V MEMORY MODULE (1x16GB) at the best online prices at eBay! Free shipping for many products!

www.ebay.com

 

[tsialex]

Btw, I've answered thinking about a single CPU tray, you have a dual one, buy at least four DIMMs. The memory I've recommended above is 1600MHz but will operate faster and cooler at 1333MHz than any 1333MHz that you can buy with a reasonable price (there are very fast and hot 1333MHz RAM with lower CL, but for exorbitant prices).

 

[sgentile92]

Btw, I've answered thinking about a single CPU tray, you have a dual one, buy at least four DIMMs. The memory I've recommended above is 1600MHz but will operate faster and cooler at 1333MHz than any 1333MHz that you can buy with a reasonable price (there are very fast and hot 1333MHz RAM with lower CL, but for exorbitant prices).

That's super valuable insight, thinking I'll buy a 4 sticks of what you suggested and if all systems are a go then spring for 2 more.

 

[tsialex]

That's super valuable insight, thinking I'll buy a 4 sticks of what you suggested and if all systems are a go then spring for 2 more.

4 16GB DIMMs will cost a little over half what you were looking before. You can get even cheaper searching around eBay, but it's best to buy from professional sellers that have a lot of exact same memory in case you need to do the RMA of one of the DIMMs or buy more later.

Btw, DDR3 price is rising hard, I'd buy six right now, the prevision for end of year is to double the current price since almost no one is making new DDR3 right now.

 

[Macsonic]

Had a client whose 2009 4,1 Mac Pro that would randomly restart. His cpus were dual 3.46 Ghz 5690s. I detached his heatsinks, applied new thermal paste and carefully reseated the 2 heatsinks. This fixed the issue and the random restarts stopped. I think his case was an improperly seated heatsink.

 

[rx78]

4 16GB DIMMs will cost a little over half what you were looking before. You can get even cheaper searching around eBay, but it's best to buy from professional sellers that have a lot of exact same memory in case you need to do the RMA of one of the DIMMs or buy more later.

Btw, DDR3 price is rising hard, I'd buy six right now, the previsions for end of year is to double the current price since almost no one is making new DDR3 right now.

Sorry to jump in but is M393B2G70BHO the same as the M393B2G70DB0-YK0 you've suggested?

They're the only ones available in my area....

 

[tsialex]

Sorry to jump in but is M393B2G70BHO the same as the M393B2G70DB0-YK0 you've suggested?

They're the only ones available in my area....

Most M393B2G70BHO DIMMs are PC10600/1333MHz, but you have to check the full part number, including the suffix.

 

[rx78]

Most M393B2G70BHO DIMMs are PC10600/1333MHz, but you have to check the full part number, including the suffix.

The sticks I'm looking at are M393B2G70BH0-YK0, so I'm assuming they are equivalent to the M393B2G70DB0-YK0?

Part numbers are way to long haha

 

[tsialex]

The sticks I'm looking at are M393B2G70BH0-YK0, so I'm assuming they are equivalent to the M393B2G70DB0-YK0?

Part numbers are way to long haha

Yes.

M393B2G70BH0-YK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70BH0-YK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

M393B2G70DB0-YK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70DB0-YK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

M393B2G70EB0-CK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70EB0-CK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

M393B2G70EB0-YK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70EB0-YK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

 

[rx78]

Yes.

M393B2G70BH0-YK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70BH0-YK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

M393B2G70DB0-YK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70DB0-YK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

M393B2G70EB0-CK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70EB0-CK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

M393B2G70EB0-YK0(DDR3) | DRAM | Samsung Semiconductor Global

RDIMM M393B2G70EB0-YK0(DDR3). Find technical product specifications, features and more at Samsung Semiconductor.

www.samsung.com

Thanks, I found the Samsung spec sheets as well.

 

[sgentile92]

4 16GB DIMMs will cost a little over half what you were looking before. You can get even cheaper searching around eBay, but it's best to buy from professional sellers that have a lot of exact same memory in case you need to do the RMA of one of the DIMMs or buy more later.

Btw, DDR3 price is rising hard, I'd buy six right now, the prevision for end of year is to double the current price since almost no one is making new DDR3 right now.

Update: Installed all 6 DIMMs and now it's crashing right as it's attempting to login after entering my user password. Tried booting from my main SSD as well as from a HDD with a fresh install of OS. I then removed 4 sticks (now running on 2 sticks) and no crash while logging in and I was able to complete a GB5 test. Thoughts? This is my 3rd set of RAM now. Could it be a problem with the way the CPUs are reading the RAM? Again, I was stable with my previous batch of 4 x 8GB OWC RAM + 2 x 2.26 processors. Over the weekend I also ran the machine with just 2 x 1GB of the original RAM and had it doing RAM intensive tasks for 48 hours straight with no crashes (but as you can imagine, completely unusable in real-world scenarios).

Am I stuck with the 2.26 processors? I'm not sure what the 2 x 3.33 Ghz processors could have introduced to the system that would cause the crashes. Thoughts?

*Update #2*
Let it sit for a while at the login screen, successfully logged into the user account on the clean-install HD, and started GB5. Got a crash halfway through. So essentially, same symptoms as with the previous batch of RAM.

 

[amstel78]

Are these new DIMMs, or are you still using the same ones from before?

 

[sgentile92]

Are these new DIMMs, or are you still using the same ones from before?

Sorry for not being clear—these are the new DIMMs.

 

[amstel78]

Ah ok.

Not sure if you mentioned this as part of your troubleshooting, but have you tried a deep clearing of NVRAM as well as resetting SMC?

Edit: after a crash, what does the following give you in terminal:

log show --predicate 'eventMessage contains "Previous shutdown cause" ' --last 10m

 

[sgentile92]

Yes I've reset both, by holding Option + Command + P + R and waiting for 3 boot chimes as well as unplugging the power cord for 30 seconds, plugging back in, waiting 15 seconds, and booting. Is there something else I should be doing?

 

[amstel78]

That should suffice, although I sometimes wait for 4th chime.

Run that terminal command. You can change 10m to 5h or whatever time frame you need. Curious to see what SMC is reporting as the crash reason.