My A8N-SLI Deluxe Saga (lesson: nforce4 sensitive to matched memory pairs)

[skipeaze]

I struggled with a very problematic situation with the A8N-SLI and a Winchester 3500+ non-overclocked.

Basically I was having intermittent crashes or CRC errors (from installers) no matter what I did. As time wore on, they increasingly would show up during the Windows XP install as "Setup was not able to copy file...." or whatever, even though I tried a variety of different (known good) Windows install CDs.

I swapped out all kinds of things to try to isolate the problems (including a second motherboard). Bought ATX v2.0 power supply, tried PCI video card (geforce 5200 series). Went through multiple everything except the CPU. Tried all the usual permutations of BIOS settings (and multiple BIOSes). Still no stability.

I finally tried something that never occured to me initially, but thank goodness I did, because I had almost given up on ever getting the A8N-SLI working.

Even though I had run memtest86 successfully on two sets of 512MB DIMMs in this system (both matched pairs and both of which I had tried separately in my quest for stability), both from Corsair, one was Value series, the other was TwinX (2.5-3-3-8), I figured, well let's pull out one of the DIMMs and see how that goes (knowing that the sock 939 processors use dual channel memory when possible). What do you know but it succeeded and everything was stable.

Went and found a third (older and slower Crucial from early '04 3-3-3-8 timings) set of 512MB DIMMs and installed those. After that perfect stability.

The bottom-line here is not that Crucial or any particular brand of memory is superior or better quality, only that for whatever reason (i have no clue why) memtest86 cannot always be relied upon to find errors. I suppose it has something to do with the A8N-SLI being very sensitive on the extent to which the "matched pairs" of DIMMs sold by memory vendors are really a true "match".

FYI (additional detail). When I was able to get the system up and running (but flaky), Everest Home Edition reported that both sticks of the problematic Corsair TwinX had the same SPD timings. Another interesting thing, either one of the sticks by themselves (system running in single channel - 512MB RAM) worked totally reliably. Also, the other set of PC3200 memory that I tried in this board (didn't work stably either in the SLI) had been running in an nforce3 board (my old system K8N-E) for 6 months, totally stable.

Hope this can help some of you out there that are having mysterious behavior with this board. Now that I've gotten beyond this, the system is a delight, totally solid, good board.

 

[Shirkin]

I have the exact same problem with a pair of Mushkin Basic's. Same exact type errors, except memtest86 will give me some erros when I have both sticks in, but none when I have only one stick in.

Guess I get to RMA and find a diff pair.

 

[Dream Operator]

I too have had issues with my memory here. At one point I was thinking RMA. I was checking around the forums at Ram Guy and got a lot of ideas for other things to check/test. One very important setting that lead to stability for me was voltage. I had tried 2.7 and 2.8 as I think those are common settings for this ram. However, I found that 2.85 did the trick for me. 2-2-2-5 no prob.

At 2.8v I was getting memtest errors with the matched pair running, on test 8. I took out one module to test them seperate and received no errors with either module. I put both in, slots reversed for fun, and failed again. I had tried 2.7 and 2.8 as I think those are common settings for this ram. However, I found that 2.85 did the trick for me. 2-2-2-5 no prob.

Not saying my answer will be your answer, I just don't feel all scenarios have been explored in the above discussion to determine the exact problem at hand.

Good luck!