RAM issues

[VirtualLarry]

I've been setting up one of my gaming rigs.
GA-P35-DS3R
4x1GB of SuperTalent DDR2-667
E2140

I have the CPU set at 410 FSB, 3280 Mhz, 1.425v in BIOS (1.36v actual).

I have the memory ratio at 2.0, meaning DDR2-820.
Ram timings 5-5-5-15, +0.2v VDIMM (around 1.984 measured by BIOS).

I first tested with 4-4-4-12, because a friend of mine has the same ram and is running at those timings, and I found an error (with Memtest86+ 1.65).

So I switched back to 5-5-5-15 timings, and I found an error at the same place, same bits. So it seems that the RAM is bad.

So I take out two of the DIMMs, and re-test, and find an error again.

So now, I've tested each of the remaining DIMM individually, and neither one showed an error.

It appears that I only seem to have an error when they are running in dual-channel mode.

Do you think that this is unusual, to only have memory fail when in dual-channel mode? The next test that I'm going to run, is to take the two DIMMs originally removed, and test them as a pair, and see if I get the same error. If I do, then it sounds like it wouldn't be the RAM, but instead, the mobo???

This is frustrating, I hate having RAM errors.

 

[Conjugal Visit]

...and if you drop your board and ram back to stock - ram at 333, 5-5-5-15, 1.8v - and run memtest in dual, you getting errors?

 

[SerpentRoyal]

Only high-quality 1.8V DDR2 667 RAMs can hit 4-4-4-12-2T with 2.0Vdimm. The average 1.8V DDR2 667 RAMs should work okay up to 360-380MHz with 5-5-5-15-2T timing.

 

[VirtualLarry]

If the RAM couldn't handle 4-4-4-12 2T, I would think that it would have given me random and assorted errors, instead of erroring in the same spot, regardless of 5-5-5-15 or 4-4-4-12.

 

[Conjugal Visit]

Other thing you might consider, Larry - and I have zero experience with Vista, or trying to run 4 sticks in a dual config - is trying to run 4 sticks of anything...Seems I've read a bunch of posts from folks that are having real difficulty trying to O/C four sticks of anything, let alone trying to O/C cheap 333 ram to 410 in a four stick config.

 

[VirtualLarry]

I wonder if bumping my GMCH voltage could have any effect.
I don't want to overcook my northbridge though.

 

[SerpentRoyal]

Originally posted by: VirtualLarry
If the RAM couldn't handle 4-4-4-12 2T, I would think that it would have given me random and assorted errors, instead of erroring in the same spot, regardless of 5-5-5-15 or 4-4-4-12.

If that particular section of the DIMM is bad, then it will fault at that same location. Retest RAMs at rated speed and voltage. There's a good chance that you will not see the error.

 

[VirtualLarry]

Ok, I took the two DIMMs that I originally removed, and tested just them in dual-channel mode, and there were no errors. That seems to rule out the mobo being the problem. So now I stuck the two DIMMs that errored back in, and am testing them again.

I wonder if the vdimm might be dipping occasionally and causing the errors? Although, again, the errors should show up randomly then, and not in the same place.

 

[SerpentRoyal]

You need reset the BIOS and test four sticks at 333MHz speed and rated voltage and timing. Some RAMs may need a little more NB and SB juice. If okay, then RAMs work as advertised.

 

[VirtualLarry]

Ok, the two DIMMs that tested bad originally, tested bad again.

Failing address: 00039a57dfc Failing bit: 00040000

So I set the BIOS back to stock DDR clocks (333 FSB, 2.0 ratio), and voltage back to stock vdimm (1.8v).

Failed again. Same place. So it's not the overclocking that did it, time to RMA I guess. I still can't explain why the DIMMs wont error out if installed singly though.

Edit: After intensive testing, singly, they still wont error. I thought that if a spot was bad, it would show up bad whether or not in dual-channel, but I guess that's not the case.

 

[taltamir]

if the same spot is always erroring, then even if it isn't defective per se (ie, it works fine in stock) it could still be a weak spot on the chip that is the first to fail always... in which case it should be replaced if you plan to do any overclocking... If it was just the overclocking then you should be having errors at random spots all over the place... not at the same spot only.

Unless a slight increase beyond that results in errors all over the place... then its just the first place to be affected. I mean if you can increase by a noticeable amount above the point where it starts erroring and its still the only spot that errors...

EDIT: Ok sorry, I seem to have missed you finding out exactly this and saying so in the post right before this one...