Mac Pro 2009 5.1 Kernel Panic

[MacZoltan]

When I run a geekbench and a heaven bench paralel this Kernel crash occurs around the end of the test, see picture.
Am i right it is one of the CPU failing to use the memory properly?
Strange as all working fine if the unit is not under stress.
Any help would be appreciated.
The motherboard and other seems working fine as I replaced the daughterboard and run the same tests successfully. Although the replacement doughterboard has only 1 CPU, while the problematic one has 2

 

[bsbeamer]

What OS version?
What USB devices connected?
What PCIe devices connected?
Using DP/mDP monitor connection or HDMI?
Have you run memtest?
Any lights inside machine?
Are you running dual benchmarks at the same time for any particular reason?

 

[MacZoltan]

second picture

What OS version?
What USB devices connected?
What PCIe devices connected?
Using DP/mDP monitor connection or HDMI?
Have you run memtest?
Any lights inside machine?
Are you running dual benchmarks at the same time for any particular reason?

All connected device and the OS is proven with many units tested exactly the same. Same specs same software.
I run paralel these benchmarks for stress testing.
This is the first time this happens.
[doublepost=1550760988][/doublepost]and the first picture

 

[startergo]

second picture


All connected device and the OS is proven with many units tested exactly the same. Same specs same software.
I run paralel these benchmarks for stress testing.
This is the first time this happens.
[doublepost=1550760988][/doublepost]and the first picture

Can you attach the report from:

/Library/Logs/DiagnosticReports

 

[tsialex]

When I run a geekbench and a heaven bench paralel this Kernel crash occurs around the end of the test, see picture.
Am i right it is one of the CPU failing to use the memory properly?
Strange as all working fine if the unit is not under stress.
Any help would be appreciated.
The motherboard and other seems working fine as I replaced the daughterboard and run the same tests successfully. Although the replacement doughterboard has only 1 CPU, while the problematic one has 2

NMI errors usually are caused by problems into QPI links. It's a 2009, processors were changed?

 

[MacZoltan]

NMI errors usually are caused by problems into QPI links. It's a 2009, processors were changed?

I did this upgrade like hundred times, and yes it is dual 3.46 hex CPU from factory 2.226 quads.
Does this QPI link error related to the daughter board or the CPU? Or even the motherboard as I did not have the chance to test the mac pro with an other dual CPU daughterboard and single CPU board requires no QPI.
Plan is to get an other unit upgrade it and if that works 100% swap the doughterboard and see what happens.

 

[tsialex]

I did this upgrade like hundred times, and yes it is dual 3.46 hex CPU from factory 2.226 quads.
Does this QPI link error related to the daughter board or the CPU?

CPU tray, can be a problem with the tray itself or the processors.

Damaged Xeons are being dumped into AliExpress by crooked sellers, some X56xx that only work with one or two channels or one of the QPI links. Some works perfectly with single trays, but don't work correctly with dual trays.

Or even the motherboard as I did not have the chance to test the mac pro with an other dual CPU daughterboard and single CPU board requires no QPI.
Plan is to get an other unit upgrade it and if that works 100% swap the doughterboard and see what happens.

Test with a dual tray that you know that correctly works, but I bet that it's a problem with the CPU tray or Xeons.

 

[MacZoltan]

CPU tray, can be a problem with the tray itself or the processors.

Damaged Xeons are being dumped into AliExpress by crooked sellers, some X56xx that only work with one or two channels or one of the QPI links. Some works perfectly with single trays, but don't work correctly with dual trays.

Test with a dual tray that you know that correctly works, but I bet that it's a problem with the CPU tray or Xeons.

my guess is too the CPU or the CPU board, I hope the CPU or one of them.
It will be a fuss anyway as I use the threadcount and screwfix liquid technique and not the delidding for upgradeing 2009s.

 

[MacZoltan]

I was too curious so I just swapped the pair of CPU with the same model but different batch and now all works fine with the MAc Pro.
This happened before and I have no clue why some CPU does not work in Mac Pro while it work just fine in a server for example.
I put the "faulty" pair of CPU in a Dell R710 with 144GB memory and the mem test and CPU test run just fine.
Strange, would be nice to find out what causing this, luckily I had only 4 CPU out of hundreds showing fault in a Mac Pro but not in any other unit.
For now this got resolved, but mistery remains

 

[tsialex]

I was too curious so I just swapped the pair of CPU with the same model but different batch and now all works fine with the MAc Pro.
This happened before and I have no clue why some CPU does not work in Mac Pro while it work just fine in a server for example.
I put the "faulty" pair of CPU in a Dell R710 with 144GB memory and the mem test and CPU test run just fine.
Strange, would be nice to find out what causing this, luckily I had only 4 CPU out of hundreds showing fault in a Mac Pro but not in any other unit.
For now this got resolved, but mistery remains

Your R710 was configured for 2 processors when you tested? I have a R410 here that I'd really love to get the second heatsink, but for around $100 on eBay…

 

[MacZoltan]

Your R710 was configured for 2 processors when you tested? I have a R410 here that I'd really love to get the second heatsink, but for around $100 on eBay…

yes, they showed up just fine and in the test section also said OK, but for these CPU you will need dual 750W psu in a R710, the dual 450W will warn you about not enough juice

 

[MacZoltan]

update, now I know that kernel panic is because of QPI for sure.
Both CPU "faulty" works just fine in a single socket unit with 64GB memory.
Passes all my stress testing without any crash.