Windows Memory Diagnostic vs. MemTest86: Which is more accurate?

[mshan]

I initially downloaded MemTest86 to test my 1 x 512 MB Buffalo Tech PC3200 (Winbond CH-5) memory module.

On my just built computer, it revealed the same failing address on almost every pass of MemTest. With continued burn-in with Prime95, the incidence of this same error has recently decreased dramatically, but not completely disappeared.

I also downloaded Windows Memory Diagnostic and just ran it overnight in extended mode with no errors reported (it reported the same failing address as Memtest86 once in all the times I've run it).

So, which is more accurate?

Asus A7N8X v2 Deluxe
AMD Athlon XP 2500+ at 166
1 x 512 MB Buffalo Tech PC3200 (2.5 3 3 8 synch at 2.7 V)
WD800JB HDD
Samsung SD-816 DVD/CD drive
Samsung FDD
Antec Sonata
Windows XP Home

[mshan]

Just popped MemTest86 into my computer after running Windows Memory Diagnostic in extended mode overnight with no errors and immediately got the same error message I always get with MemTest86:

test 4
pass (first 3 passes led to 3 error messages)
failing address 000009883ed4-152.1 MB
good fffffffb
bad ffffffdb
Err-bits 00000020
count (this time it says 2 1)
chan

This is with the standard tests. When I run all tests, the same error message shows up in test 11 and Err-bits says 000000.

What's going on? Is my memory bad or is this just a false positive under MemTest86?

(I have increased VDIMM to 2.8V in the past, and this has resulted in an even lower incidence of error messages in MemTest86 - 1 error in overnight testing - but I would like to use the lowest VDIMM necessary to produce error free function of my computer so as to preserve the longevity of my memory module. Is there any significant difference in heat production between VDIMM 2.7 and VDIMM 2.8, either for the memory module or motherboard?)

[notdrugged]

You're not the only one to notice one or the other catching small errors. With as little incidence, you may not notice much, or you may have data corruption down the road, it's hard to tell.

Trust both memory diagnostic programs. They don't lie. You should try to figure out why the errors are happening. I'm thinking anywhere from bios settings to hardware trouble.

I also intended to mention that WMD has extended tests also.

[mshan]

In interpreting these two memory tests, what does the fact that:

1) it is always this one memory address that fails
2) the incidence seems to be decreasing with continued burn-in and possibly increased VDIMM (2.7V)

mean?

Is it more likely to be a correctable system incompatability / improper setting, or a physical defect in the memory module that I won't be able to correct without replacement?