I am completely confused with the latest hurdle from my machine.  Whenever I run my ram in dual channel mode I get random restarts... memory dumps and such, so I run my system through memtest86 and the errors show up like it's a party.  So I remove one dimm, test it, still errors.  I switched the dimms (always using slot0 on my motherboard) and still errors.  However, to run dual channel I need slot0 + slot1.  Well... when I run both dimms in slot1 and slot2, single channel, not a single error, stable as a rock.... I'm confused.  I overclock, but not once did I raise the FSB on my board for 400mhz, nor did I ever apply tighter timings or overclock my memory.. I don't get it?  I don't know if this problem existed from day1 so I'm thinking it was possible to get a faulty motherboard?  Any input would be greatly appreciated..  
 
Abit NF7-S 2.0 (latest bios)
Zalman 7000cu (modded to fit)
AMD 2500+ XP-M 12x200 @ 1.75 volt
OCZ 512MB X2 DDR400 2-3-3-6  
Soundblaster X-Fi Music
BFG 7800gs 460/13.5
Cooler Master 380watt PSU
WD 250gb 16mb SATAII x2 raid0

I would suspect the motherboard - that socket sounds blown.
 
Try blowing out the slot with canned air, make sure no dust or anything blocks the DIMMs from socketing fully, and if it still errors in slot 0, RMA or replace the board (or live with single-channel in slots 1&2).
 
Mike.

You should test the memory on a different system to better isolate the problem.

The problem is the motherboard not the RAM, I have the same problem with my cousin's motherboard it will not run Duel channel, if I run duel channel mode the computer will restart randomly. If I put the RAM only slot 1 and slot 2 everything works fine, not a single problem. But I didn't bother RMA'ing the motherboard... just live with single channel, not like my cousin knows anything about it LOL

Ya, and dual channel vs single channel isn't a huge performance difference anyway, I'm just glad I got it stable.  Thanks everyone for your input.

Your post is a little confusing.  In order to run the NF7-S V2.0 (right?? not the S2 version) in dual channel, it is recommended to run one chip in DIMM3 and one in either DIMM 2 or 1.  This doesn't concur with what you posted, unless you're getting the DIMM slots backwards.  DIMM 3 is the slot by itself.  Install your memory chips from that slot down, i.e. 3, 2, 1.
 
You said that "not once did I raise the FSB on my board for 400mhz" and yet you're running your mobile at a 200 MHZ bus, well above it's 133 MHZ bus.  This could be part of the problem.  However, how much voltage are you supplying the RAM and have you tried loosening the timings?  I don't care what OCZ says the timings should be.  You have to run it at what it works at.  I would start there to see if it helps.
 
I'm not sure you can RMA the board since it's so old (but one of the best).  BTW, if you get the chance, what kind of temps are you getting?

According to the manual the DIMMs are named DIMM0, DIMM1, DIMM2, starting from the outside of the board (the DIMM by itself) is DIMM0.  I have tried different memory configurations as well, while in dualchannel.  Also loosening timings did not help either.  My cpu temp at 100% load is 54 degrees.

Infact if the timings are too loose (8-4-4-3) my system wont boot, but thats not a problem, but it is a bit odd.

That's interesting, because my manual notes the exact opposite.  However, it seems as though you are starting at the bank by itself and moving down from there when installing your dimms which is correct for dual channel to run.
 
Before narrowing it down to the motherboard, is there a way you can test the dimms in another dual channel system like DrBlofeld suggested to see if it recreates the errors in dual channel?  Also, what voltage are you supplying the RAM?

I might be able to test dual channel in another system, we'll have to see.  As far as voltage, I left it at default, 2.6.  If your curious of the exact ram I am using here is a link.
http://www.ocztechnology.com/produ [...] el_gold_c2

Hmm.  I know you can at least get away with bumping the voltage up to 2.8 without any worries.  You might want to try that.  According to the link you posted, you can bump your voltage up to 3.0 without invalidating the warranty.
 
I know the RAM I used in my NF7 didn't run at DDR400 speed with the advertised timings of 2.5-4-4-8 without failing memmtest, but it was just cheap Geil Value RAM.  After bumping the voltage up to up to 2.8 I was able to get it to pass memtest on 3-4-4-6, IIRC.  It took a lot of minute adjustments to the timings as well as patience to finally get it to pass memtest without errors.  It still runs rock solid, just not with lightning fast timings.
 
This, and testing in another system are worth a shot before writing off the best overclocking board for Socket A.[/fanboyism]
 
BTW, did you have to replace your northbridge fan too??  This was the only flaw with most of the NF7-S V2.0 boards.

Yes I did, with the Zalman 7000cu it works quite nicely too.  I have one of these...
http://www.zalman.co.kr/eng/produc [...] 1&code=014
With some of the fins hacked off with a dremel so it fits with that massive 7000cu cooler beside it.  I don't know exactly my northbridge temps but they stay pretty low, I think between 35-40 if I'm not mistaken, and I'll give the voltage jump a try.  Thanks

Nice and passive.  I went the Vantec CCB-A1C route myself.  Used the extra heatsinks to cool the southbridge too.
 
Good luck on the voltage.

Problems like that are common with agressive memory controllers (like the nForce2 and 875), especially when the memory is crap.  OCZ has made lots of crap RAM, most of their RAM requires raised voltage to run at rated timings, and this is even more important with boards like yours.
 
Try 2.8v.