Machine check exception - how to read and understand it?

Question

My computer acts strange - sometimes reboots spontanousely, sometimes does not want to boot. When I isse command "mcelog", I receive such info:

HARDWARE ERROR. This is *NOT* a software problem!
Please contact your hardware vendor
MCE 0
CPU 0 BANK 0 
MISC 61f80000500010e ADDR cf9 
TIME 1305836650 Thu May 19 22:24:10 2011
MCG status:
MCi status:
Uncorrected error
MCi_MISC register valid
MCi_ADDR register valid
Processor context corrupt
MCA: BUS Level-3 Observed-error-as-third-party Generic Memory-access     Request-did-not-timeout Error
Model:Response hard fail
STATUS ae00030010020c03 MCGSTATUS 0
MCGCAP c0204 APICID 0 SOCKETID 0 
CPUID Vendor Intel Family 15 Model 4

How can I read and understand it? Could somebody point me to some documentation?

Answer 1

I've been looking into MCE errors because I'm currently working on tuning the overclocking on my machine. I haven't run into any, as I caught all the bad settings with MemTest86+ first, but I know I might, so I have to regularly check for them for the first few months after an overclocking change. They can crop up when you have the hardware pushed just barely past it's limit, and are a sign you overclocked a tad too far.

The following lines are important

"HARDWARE ERROR."

and "MCA: BUS Level-3 Observed-error-as-third-party Generic Memory-access Request-did-not-timeout Error Model:Response hard fail"

The other lines say the registers in the processor are not causing the failure, and the exact error specs and exactly what was impacted by the error. You aren't going to need any of this info unless you are a kernel developer or a motherboard developer.

It appears your error is from the memory. It is what is commonly referred to as a die-hard fail, because your system is booting, just getting errors.

The following are common causes for issues with memory, memory controller or bus. (In order of the ease to fix the issue._

1. Overclocking issues. (Timings on RAM are too short, RAM bus speed is too high.)

2. Voltage issues (Voltage to RAM and/or CPU is set wrong in BIOS, too low or too high or board is designed for a different voltage RAM, e.g. you put 1.65 volt RAM in a board that takes 1.5 volt RAM.)

3. Overheat issues (CPU RAM controller, CPU Cache, motherboard and/or RAM is overheating. This may have to do with voltage issues.)

4. Bad power supply (This is due to big issues.)

5. Bad Memory (Try testing with MemTest86+ including the dreaded bitfade test. It may not be the memory even if detected.)

   6 Bad BIOS (WARNING, it may be dangerous to flash the BIOS while you have bad memory. Check your manufacturer's website to see if there are issues causing memory corruption, download and prepare the BIOS image on a different computer and use the on boot BIOS flasher, to minimize amount of resources in use, and thus the amount of things that can go wrong.)

6. Bad motherboard and/or bad CPU. (I think this is obvious.)

Answer 2

Only line that counts,

Uncorrected error

you have either bad memory or a bad motherboard. Swap one of the two and try again. Run Memtest86 to find if the error is still there faster and possibly try running memory at a slower clock speed.

Unless if you're going to hand solder replacement chips, nothing else really matters. Not that I'd recommend doing that anyway.

Answer 3

The Intel (or AMD) system programming manuals are once place to go for MCE decoding help (look at the "Machine Check Architecture" chapter). They're pretty dense documents, though. And you'll quickly get into the weeds of micro-architecture revision differences, chipset vs. CPU issues, etc.