Is it RAM or memory controller problem?

[Brad Williams]

Got a bit of an interesting problem here... webserver (running Win2k Server) throws STOP errors quite often (Mostly IRQL_NOT_LESS_OR_EQUAL or KMODE_EXCEPTION_NOT_HANDLED). Repair of the O/S did nothing, nor did a clean reinstall.

Decided to run Memtest86, and I'm getting large blocks of errors in which the data written to the memory is the inverse of the expected data (IE: expected FFFFEFFF, memory contains 0000100). I suspected the memory, but if the wrong data is being written, I considered the memory controller to be at fault.

Any ideas?

System Specs:
Tyan Thunder S2882 Motherboard
1x Opteron 265 Processor
2GB Kingston HyperX ECC Registered DDR400 RAM (4x512MB).
WD Raptor 74GB 10kRPM SATA HDD
650W EPS12v PSU (unsure of brand without opening up the server, but voltages within range)

EDIT: Memtest86 also detected something along the lines of 49000 ECC errors.

EDIT 2: Ran Memtest without the ECC option on, and it's not reporting any errors (neither the wrong-data errors nor ECC errors).

 

[Tedster]

any errors in memtest are not good. swap out sticks one at a time until you find the faulty ones. In rare cases you can have a bad slot. (unlikely though)
Overclocking will cause errors and destroy RAM.
You must use the correct ECC type for your memory. Check with the manufactuer. Memtest will give false answers if you mark the wrong type.

 

[zephead]

i would stick to corsair memory.

 

[Brad Williams]

Thanks for the reply, Tedster. The problem was indeed a bad pair of sticks, so once those got replaced with new ones, everything worked fine.

Thanks again!

 

[Tedster]

glad to hear it was resolved.