strange memtest86 error on i875 platform

[klenot]

Hi everybody,

While running memtest86 v3.1 on i875P based mobo (http://www.fic.com.tw/product/motherboard/intel/p4-875pmax.aspx) settled by two pieces of 256MB Hyperx (BH-5) modules, i got errors in test #6 (all other test run fine).

FSB: 218 MHz (CPU can handle it perfectly stable)
Timings: 2-2-2-5, lowering to 2-3-2-6 did not helped.
VDIMM 2.8V

The strange thing is that good pattern should be ffffffff, and the bad (which memory returned) is ... also ffffffff. So, .. what's wrong? Why memtest detected it as error if both expected and real value are the same?

http://andilek.com/ocforumspix/memtest86v3.1-test6-i875-strange-error.jpg

[RedDragonXXX]

Try getting the 3.2 Version and see if it does the same thing.

[klenot]

It has been caused by faulty CPU.

Tried changing everything, one can imagine, except the CPU. No success. As last resort, I changed the CPU. Problem disappeared immediately. Putting the damaged CPU back restored this strange errors.

Curious thing is, that when I booted into Windows 98 (with this faulty CPU), and run Prime 95, it was running stable for dozens of hours, even overclocked by 9% (at FSB 218 MHz, default voltage).

Strange.

Anyhow, my advice is not torturing the CPU with too high voltages. I don't know what voltage caused this damage (it was not me who did it), but I suspect it was something well over 1.65 V (The CPU core is Northwood D1).

I did never seen a partially working CPU. Well.. now I know, that it is possible. So, testing CPU the way "it POSTed, so it is surely completely faultless" is an insufficient way of CPU testing.

[X0d1@k]

THe wierdest thing I ever saw Sisandra do was give me a current CPU temp of 476 degrees F. Needless to say it wasnt that hot, or close to it, but someone recommended that I flash a new BIOS on my motherboard.