I finished my new build with:
Gigabyte-P35-DS3L
E8400 CPU
4 x 1 GB Crucial Ballistix Memory PC 6400 (from mobo list)
Windows Vista Ultimate

I finished the build and everything seems to be working fine with low temps. Then I ran memtest=86 v2.01 and get errors? Now what?

During the build I first installed one memory stick. After successfully booting windows, I increased the RAM voltage from the mobo default of 1.8v to the memory spec of 2.2 and installed the other 3 sticks. I then contiued with other tweaking, loading some software and using the system.

When I loaded CPUz, I read that memory settings were 5-5-5-18. It reported 4 possible alternate settings from SPD, including 5-5-5-18 at 1.8v - what it was running at - and 4-4-4-12 at 2.2v. I had hoped that when I increased the voltage to 2.2v it would let the memory work at the faster setting - but apparently it did not.

I then when into the BIOS and changed the timings to the second, faster set - changing only those particular settings - plus changing the related control line from auto to manual. Everything seems to still work fine and CPUz now reports the faster timings.

So I ran memtest for the first time and it reported one error on the first pass. It now has run for over 3 hours, is in the fifth pass, and reports 6 errors.

In the errors section, only one error is listed. I was thinking this meant it keep finding an error at the same location. But then I noticed on the bottom that scroll was locked. Now there is still half a page on which it can list errors without scrolling, but does scroll lock mean it will only list the first?

I tuned off scroll lock and will see if any more get listed. But then what do I do?

Change the timings back and repeat the test?
Start testing each stick to see if I can isolate the problem? If I can to one stick, does that likely mean the stick is bad and I RMA it?
Keep running the test to see if there are any other meaningful results?
Do something else?

Welcome to reality. Rated specs aren't guaranteed to work on EVERY config.

I suspect your ram isn't micron d9. Mine can do 4-4-4-10@2.1v. I'm not getting rid of my 4 sticks anytime since I hear the new ones are crap chips.

You can RMA it, but you'll probably get the same result with the new sticks. Or you can loosen up the timings and live with that. There's a very little difference between the C5 & C4.

Crucial traded out the D9's for some crappy single sided chips on their more recent RAM and then lowered the price to make it look like a good deal, 2x1gb DDR2 800 can be had for $20 after MIR. RMA it until you get a set that runs as it should. Eventually Crucial might get the point and not do that bait and switch thing again.

At my store- we run memtest on all RAM before we send out builds to customers. If Memtest even detects one error we replace the RAM.

Try it again at 5-5-5-15 and see what happens. I suspect the error will still show up however.

Only one error it is probably not worth the hassle of sending the RAM back and waiting for new stuff. Remember it has a lifetime warranty on it- so if it gets worse just get some new stuff from the manufacture.

Since I was seeing a single, repeated error on almost every pass, I thought the error would quickly show and repeat itself with when testing the errant stick. But I ran Memtest on each stick indvidually for 4 passes and it reported no errors for any stick.

I reinstalled all sticks but in reverse order, i.e. 1-2-3-4 became 4-3-2-1. This placed each stick in a different channel and reversed the order of the two sticks in each channel, and ran Memtest. The result is 2 errors in 14 passes at an address similar but not equal to the prior address:

Trial Passes Errors Failing address Good Bad Err-Bits
#1 24 22 000beaea4c0 - 3050.6MB 00000001 0000021 00000020
#2 14 2 000be8ea480 - 3048.6MB 00000001 0000021 00000020

If the problem is on one stick of memory and I moved all sticks, should not this change the address of the error by at least 1 MB? Since it apparently did not, does this indicate a mobo problem? Or am I misinterpreting the test results? Or are there any other BIOS settings that should have been changed along with the specific timing settings?

It would apper you might have a faulty RAM slot on your motherboard.

I am not sure what problem I have.

I tried one stick of RAM in each slot for 4 passes. 3 Slots passed all passes but slot #3 had 8 errors in 4 passes. So I figured I had a mobo problem,
Then just to followup on another suggestion above, I changed the timings back from the 4-4-4-12 (which as you recall is one of the listed settings for this chip and the chip is on the mobo approved list) to the original 5-5-5-18 as automatically selected originally by the mobo. 4 passes and no errors. Maybe I just had to give up the faster timings.

I then installed all of the chips back in the mobo and ran Memtest again, hoping this cleared the problem. I ran it for over 5 years and completed 9 passes - but with 28 errors - a much higher rate than originally at the faster timings. Now when I changed the timings back to 5-5-5-18, I did not change the voltage from the 2.2 spec for the 4-4-4-12 to the 1.8 spec for 5-5-5-12. Could this be a problem? Why would the 5-5-5-18 work better in the one slot, but worse when occupying all 4 slots?

How much slower is the 5-5-5-18? One posting above said it was not much slower but I had read that the actual bandwidth was directly proportional to the timings, and that 5-5-5 would essentially be 25% slower than 4-4-4.

In any case, instead of trying that I ran Prime95. I ran it for about an hour at the blended setting (which it said would put significant stress on the memory) with no errors. I changed the test settings to select larger unit sizes (1750 KB to 2400 KB) to further test memory. I ran it another hour with no errors. I changed the memory timings back to 4-4-4-12 and ran it again on the blended setting. Again no errors. I have still have had no problems in regular use - although I have had limited use while it was tied up running tests - I still use my old PC for most things since I have not completed the setup and have not transferred files over.

Do I have a mobo problem or what? I hope not as I don't really want to change out the mobo and essentially have to rebuild. Also because I may have difficulty getting an exact replacement GA-P35-DS3L since Gigabyte is replacing those with a new board. I purchased the board at Fry's and stopped by one store here in San Jose to check availabililty and all they had left were open box units.

Am I correct in assuming if I replace it with the same board I will not have to reinstall Vista, but if I replace it with another board, including the newer GA-EP35-DS3L I will need to reformat, or can I just change the mobo drivers?

I am hoping I can finishing setting up my computer (including loading various application software) over this long weekend - but I am not sure in what direction to go next.

My system is now stable and passes repeated Memtest runs, albeit with slower memory.

Another poster on THG Forums suggested that anyone with Gigabyte mobo issues go to Tweaktown.com where they have a Gigabyte forum hosted by a Gigabyte moderator - an excellent suggestion. I first searched the posts and found some tweaks to try, but also by reviewing posts and responses from the moderator, lsdmeaculp, I was able to better understand and frame my problem. When I posted, I reported what I had already done based on his suggestions to others and the data I knew he needed to look at. Within a few hours - on July 4th - a holiday no less - he posted his suggestions - which was all I needed to make my system stable.

To make the Crucial memory work on the Gigabyte board, I had to change some of the other timing delays in the BIOS. I was able to keep the settings at 4-4-4-12 but my memory is slower because of the other timing changes. Now I am in the process of slowly tightening them to see how fast I can get the memory back to and still be stable.

I have my BIOS settings for memory back to "normal" and my memory is running fine at 4-4-4-12.

For others that may be having problems with the Crucial Ballistix 800MHz memory on the GA-P35-DS3L board, it appears that the key memory change needed was to loosen the Rank Write to Read Delay setting from 3 to about 8. I took it back down to 6, but my benchmark tests reported a slower result than at 8, which the expert at Tweaktown said was not surprising.

Note that before even getting into working the timing issues I had increased the DDR2 OverVoltage +0.3 to (making the memory voltage 2.1 compared to manufacturers recommended default of 2.2v) and changed the general timngs to 4-4-4-12. The rest comes after that.

I changed several other timings - taking them first off auto and raising em then slowly lowering them back. These are the final settings (in addition to the standard 4-4-4-12).

ACT to ACT Delay (tRRD)_____ 3
Rank Write to READ Delay____ 8
Write to Precharge Delay_____ 5
Refresh to ACT Delay________ 52
Read to Precharge Delay_____ 3
Static tRead Value__________ 6

I also changed these when it was recommended for general stability:
FSB OverVoltage Control +0.1
(G)MCH OverVoltage Control +0.1

I hope this helps others avoid all the time I spend jumping my timings up then cranking them down one step at a time.