Odd results with Uwackme´s bh-5 burn in

[Guest twilight]

I read the post in this forum about Uwackme´s bh-5 burn in and decided to try that with my bh-5 sticks but my results are´n going the right way....

 

9*255 fsb

CPC ON

5-2-2-2-2.0-9-12

the seven other timings AUTO

Active cooling on NB/SB

CPU Temp ~ 55c

Vcore 1.9

Vdimm 3.3

AGP Volt AUTO

cant remember the name for NB but at 1.9 volt

 

 

First 6 hours ~ 200 error

 

next 6 hours ~ 4800 error :confused:

 

last 3 hours ~ 4000 error (Aborted test) :mad:

 

I has hoped that it would be less errors the longer i ran the test but no......look at my horrible numbers above :eek:

 

Even tried 5 min more but got 1000 errors in that time so.....:sad:

 

Almost all my error is between 480 Mb - 508 mb in the error log.

 

Any ideas what can be the reason for theese accelerated errors?

Are my BH-5 sticks ready for the trash can, or does my NB or CPU belong there instead?

 

Another thing, i ran the test with both sticks in dual mode can that be the cause?

[wez]

I would say that your RAM got so hot that it started to write errors.

Try to have a fan blowing on tha RAMs when you do the burn-in.

[Guest twilight]

I had a 92 mm enermax fan blowing at the mems, but they maybe got to hot anyway.....maybe thats the reason for my overall unstability with my computer....it works well for the first couple of 3d mark tests then the trouble begins....the same for gaming, 30 min - 1hour then a sudden reboot or something strange...

[wez]

Or maby your NB is getting to hot to, you have the std HS on it?

:confused:

 

edit: didnt see that you had active cooling on the NB.

[SAE]

No, that effect is what I am experiencing at 268-272MHz max timings in memtest

 

It seems the chips really get too hot to run errorfree... I dunno, but maybe some biga$$ sinks attached to the bh-5 with good heat-transferring adhesive/tape can reduce the errors.

 

Only problem is: How to get biga$$ sinks onto the dimms and run 2modules at a time!? I think there must be a compromise... maybe a flat waterblock...

 

My NB is lapped and watercooled.

[wez]

Yea, i got the same problem when iam didnt had extra cooling when i did the burn-in, but after that i had a fan blowing on the mem the errors disapered.

[Guest twilight]

My active cooling on the NB is probably not so good....its the original heatsink with a fan mouented on it, but i have a feeling that the original heatsink doesent pick up the heat from the NB because its never feels hot if i touch it, even without the fan. Maybe i should change heatsink on the NB and make sure there is a good connection surface between the NB and HS.

[uwackme]

One thing to do is back off the FSB to the point where errors are 0.... tell us what that FSB is.

 

Run there for a few hours then see if inching UP in FSB again yeilds errors right off or what.

 

Cooling the rams is important, just tiewrap an 80mm fan directly over the DIMM during the test.....I tiewrap it to the powerwire of my CDdrive as a quick attachment point.

 

The NB can be very "concave" and need a minor lapping to get a good contact with the hsink. I now 1000grit lap my NB's always and reapply compound....thin coat of AS3 ceramique...always. HS not getting warm might be from totally crappy contact with NB...check it.

 

Also try backing off .1V to Vdimm 3.2V just incase the regulators are at thier max for some reason.

 

Also, there is always the fact that the ram wil hit an absolute wall at some FSB.... once you hit it, thats that, you found the max OC.

 

 

Can also be a situation where your ability to do CPC ON is on the hairy edge and just enough heat/FSB pushes it into errors zone HARD. Try testing for max with CPC off and see if behavior changes.

[Tedy]

can you make a pic how you're cooling ram ?

 

found 1 way

 

lol

 

http://forums.extremeoverclocking.com/show...ad.php?t=113513

 

Required:

 

1: A good fan (preferably 60x60mm, stock Amd heatsink fan for instance)

2: Fan screws (large diameter)

3: Two elastic bands

 

 

[Guest twilight]

Ok....i will try to lower the volt vdimm to 3.2 and trie some lower bus settings, cpc off and see what happens....

 

Thanx for all the help guys

[RGone]

...but if you go back to the 'first' place where 'uwackme' started posting the BURN-IN> he said "72 hours" and "not" 3 hours now; 6 later and 2 hours the next and that sort of burn-in arrangement. You ask him and he will tell you so. Most cried back then from having to think of 72 hours without puter and 'uwackme' said he putered on another rig and or burned in on another rig. Everybody is always so persistent in shortcutting the orginal finders procedure. It is the shortcutting and hurrying or whatever that makes for poor results on most anything.

 

Now I know 'uwackme' has replied in this thread and he made a number of suggestions as have all. Some of them perhaps helpful and some might point to you are truly topped out. But when 'uwackme' hooked me on this BURN-IN thing it was impressed on me and those at the time that you have to RUN it more than 6 hours and 3 hours at a time. Why? I have posted the why once or twice in some old threads. Burns must go longer than 6 hours because over the course of 8 sticks I did with good results the memory would cycle thru X amount of errors up thru about 4-5 hours and then *really* start to put out the errors. Bunches of errors. Tons of errors. But after 5-10 hours of CONTINUOUS running the errors would begin to drop. You have to run it until the errors turn the corner and start back down to see results most of the time. So my first burn is always over 20 hours so that the memory gets a chance to cycle up in errors and then down in order to see any real results.

 

Now out of 8 sticks of memory two of which were CH-5> I had 4 sticks produce excellent results. 2x256 CH-5 went from 243FSB performers to 252FSB performers. 2x256 BH-5 went from 248FSB performers to 255FSB performers. The other 4 sticks gave mixed results but 'lost' no speed due to errors.

 

One must note and remember that YMMV but the orginal postulation set forth by 'uwackme' was for his 72 hour continuous burn in. Now that burn-in very likely does work. All other procedures based on my own personal observations may wildly vary. It is the cycling that is seen and gone thru until the downhill trend of errors comes and BURN-IN continues that does the deed.

 

Sincerely, RGone...

[Guest twilight]

Ok....The 6 hour + 6 hour + 3 hour periods i mentioned was done with maybe 10 seconds between each period, i just restarted the test to see if the result would continue to be the same.

 

I know that it said 72 hour in the burn in process but i didn know that my error should increase first , then decrease after many many hours....i thougt that i would have an steady amount of errors from the beginning, then maybe after 40-50 hours the errors would start to drop.

 

It was the enormous acceleration that scared me a bit, i was afraid that i should toast my precious bh-5 sticks or my NB. I will try it again and try to NOT turn off the test when my memorys beg for mercy

 

just a little question: how many errors per hour is a good starting point for the first 10-15 hours? I mean the whole effect of the burn in must be destroyd if there is to many errors, right?

 

Thanx for your time and replies