new build: RAM fails memtest 5&6 but only in dual-channel

[Kochab]

I put my new system together, set everything up on default and checked for stability (Prime95 and bootable memtest86 v3.2). Everything was fine. Then I got bored and I messed around in with the voltages and such. That's when I noticed that my RAM was getting errors. I returned all my settings to default, even resetting the BIOS to be sure, but I still got 1 or 2 errors on tests 5 or 6. I missed this earlier because the errors generally occur after several successful runs (I was paranoid that my memory would get errors running at anything other than stock so I let memtest run for much longer). Narrowing things down I removed one stick, had 14 successful runs than checked the other stick for 14 successful runs. I put both back in and I get the same old problem. After this I did a Prime95 run (on each core) utilizing most of my RAM, and got over 14 hours with no errors or warnings.

So now I'm thinking it has something to do with A) my mainboard (which, by the way, was squashed in shipping) or B) the memory controller on my AMD dual-core, or C) I just found something here on Anandtech:forums about somebody using an ASUS A8N32-SLI and getting the same errors caused by a USB setting, or D) the PSU, or E) the memtest program itself (I've heard it doesn't make provisions for dual-cores)

here's my relevant hardware:

mainboard: Asus A8N-SLI Premium
CPU: AMD athlon 2.2GHz X2 512KB L2 per core
memory: Kingston DDR400 ECC 2x1GB
PSU: Coolmax CUG-600B 600W with modular cabling
video card: ATI All-in-Wonder X1900

UPDATE: did a little more research on the USB legacy support / mainboard problem. I had already disabled it, so we can rule that out.

FINAL UPDATE: Changed the command rate from 1T to 2T. Problem solved.

 

[jiffylube1024]

What voltage are you giving the RAM? 2.5V? 2.6V? Whatever it is, bump it up 0.1V and see if that fixes the problem.

 

[Kochab]

....tried it already. tried 2.6 incrementally up to 2.8, no changes.

 

[Kochab]

this may or may not have anything to do with my symptoms but when I tried overclocking with the above components I was very disappointed.

here are my results:

HTT: 227*5
CPU: 226*11@1.375v (LDBFE 0623)
MEM: 211@2.750v

...and yes, each overlock was isolated from the others.

that just seems awfully weak. again my suspicions lie with my mainboard, being squashed an' all, damn you UPS apes! damn you to hell!. I should have refused the package right there.

 

[GeezerMan]

MY Asus A8N-E board, with 3800+ X2, Corsair XMS 2GB does odd things too. With the sticks in a certain order, I get mem test errors. By themselves, I got no errors. I checked with the Ramguy over at the Corsair forum and he said to swap the order around
for the sticks. It passes memtest now. So, flip your sticks around.

 

[myocardia]

You have to shoot for a maximum of 1000 total on the HTT, not a minimum.

 

[Kochab]

Geezerman, ASUS recommends using slot B1 when using one stick, so I've inadvertantly changed the order already, since I tested both sticks. So it seems it's some design flaw in the board itself, ugh.