burn in to long, errors return?

[Da_BoKa]

Today i wanted to perform a descent burn in on my twinmos bh5 modules.
I've set the vdimm up to 3.3v ( measured by DMM ) and upped the fsb until 255 at cl2 2-2-5! offcourse it generated some errors ( always 2 ), not at each loop but in the beginning every ~10 loops.
But then the gaps between the passed tests and the errors started to grow, from 10 passes to 20 and so on until after 6 hours ( less than 100 errors in 8 hours ), it took up to 100 passes to generate another error ( yet still at the same adress ) and it ran up to more than 300 passes to generate another error ( this whas after 8-9 hours ). So i tought just a couple off hours more and the errors would be completely gone, but no! When i returned to my pc i saw that the errors where back and this time the gap was +-100 passed tests ( offcourse #5 ), this whas after 14 hours of burning in!
I"ve immediately shutted the pc down to let the mem rest!
Is this normal that the errors return?

It was cooled by a 80mm enermax fan the whole time!

[justwOo]

14 hours of burning in on bh5 is childs play
Theres no way that hurt your ram if you had it at 3.3volts and a fan directly blowing on the ram
I say back it down to 250-253 and try again and go from there
Did you test the volts going to teh ram while it was looping?
You could also try leaving it at 255 and try giving it 3.35-3.4 and see what happens

Its all about playing around with settings... maybe even try 2-2-2-6

[Da_BoKa]

Yes, i monitored the voltage and it was completely stable!
I just restarted the burn in at the same settings and almost 300 passes without a single error, i hope they are gone, so i can up the fsb a little bit again!
I know the mem can do 260 on 3.35v super pi stable, haven't tried memtest tough.

[cantankerous]

Sounds like you have some nice sticks there. Wish mine would do that.

[Da_BoKa]

400+ passes and no errors, it seems like they are gone.
still can't figure out why they came back yesterday!
burning atm at 258 mhz at 3.35v :)