i5 won't boot with all cores enabled

[yongleflibbit]

Hello everyone,

Sorry for the long post, but it's complicated, and sort of interesting.

I'm trying to diagnose which of my components is faulty/defect/damaged, and possible RMA.

Story so far:

Got a P67A-GD65(B3), i5-2500K. Was running fine for a week with 560 Ti (Frozr II).

Then decided to try OC genie. Worked great at 4.2. Used the same voltages and manually changed to 4.6 (46x100), ran for half a day.
Then later, from the desktop, the PC shut down without an error.
On reboot, it POSTed OK but bluescreened when starting Windows 7 with this error:
'MULTIPROCESSOR_CONFIGURATION_NOT_SUPPORTED'
Startup Repair wouldn't start either - same error.

I tried:
- Reset the BIOS (switch, jumper and battery removal)
- 1 stick of RAM, all other items unplugged, remove/replace ATX connectors and CPU
- XMP on/off
Single core always started, but never 2 or more.

So I thought, OK, I'll try single core mode. Windows booted OK - everything worked fine, games, surfing, just a bit slow. Activated XMP - 1600 MHz no problem. Even Sli with a second Frozr.
Next I tried 3 cores: failed. 2 cores: failed. Tried booting Linux from USB to test if it was a software problem - it never booted whatever cores I picked.

Opened a ticket with MSI Support - they said to install the Intel Management Engine - I had this installed already, so reinstalled. No change.
I decided I had killed my CPU, and got some money to replace it.

A few days later, I put a new i5 2500k in. Powered on OK.
Got a BIOS message saying CPU/memory changed - I chose Reset to defaults.
The board power cycled: The first and second BIOSes failed(slow then fast flashing). No display.

Tried the above steps again - the board would not POST with default settings. BIOS warning flashing, but I couldn't get in to re-flash it.

Then I replaced the old CPU.
I could now POST and enter BIOS no problem. Windows would still only start in single core mode. Used the windows .exe file to flash to the same version I was using (1.3).

I flashed to 1.7(in BIOS): no difference.
Flashed to 1B(in BIOS): no difference.

I then put the new CPU back in - It POSTed! Windows also OK!
2 cores: Still POSTed! Windows also OK!
3 cores: back to power cycling (but no BIOS fail this time) Had to swap back to old, change setting to 1 again, and back to new.

Right at this moment I am running with the new CPU with 2 active cores. No heavy apps have been run, no OC Genie.
The RAM(G-Skill 1600 CL9, type ends in BNQ, on QVL), PSU (a Corsair TX750W), Alpenf?hn cooler and both Frozrs all come from my old PC and have run fine.

So:
The board and both CPUs are both happy to start Windows in 1-core mode.
1 (newer batch + serial) CPU won't POST in 3 or 4 core mode, but will POST and start windows with 2 cores enabled.

Now there are lots of questions:
- Which part is faulty?
- Is the motherboard really OK? (power phase control? Mosfets dead? c-state or multi CPU control burned out?)
- If not, is it likely this could damage the new CPU?
- What about RMA? (I am already planning to buy a new P67A-GD65 anyway because I love the board!)

Any and all advice appreciated. (I have images and serial/batch numbers, I will post them later if anyone thinks it will help)

 

[HU16E]

Recommend removing the CPU & with a strong light & magnifier, checking the pad to pin contact marks on the CPU pads, & inspecting the CPU socket pins very closely for any that may be broken or bent. Uneven torque pressure from an aftermarket CPU cooler can cause odd behavior as well.

 

[yongleflibbit]

I inspected both CPUs, the pads all have marks where the pins make contact. Almost all are neat dots, some not so easy to see, some are very short lines where the pin has moved a little, probably as the CPU is locked into place.
The socket looks perfect, all pins are neatly aligned.

About torque: I removed the Alpenf?hn completely and used the stock Intel cooler, to see if the pressure was causing weak contact. Nothing changed.
The pin contact is probably OK anyway - it was fine when I first set up the board.


I'm back to 2 cores on the newer CPU. Uniengine benchmark completed OK, also in SLI. Then tried OC Genie: Ran a Uniengine and a Crysis benchmark - Everything seems OK apart from the number of cores.
This problem is getting very difficult to track down!

I'm now going to try the other settings related to cores: the ratio limits. I'll set them all to 37 (the default 1 core limit) and see if it will run that way - it probably won't, so after that I'll have to swap CPUs in and out again and won't be able to post for a while...wish me luck!