uwackme's bh-5 burn-in

[TheWizzard]

I have another question: is it possible to mix, for example, a memory module of corsair and ocz, both bh-5 512mb, to run it in dual channel?

[Emil]

Both being BH-5, the chances of a good mix is better than fair.

The rule of thumb however would say that even with the same brand and chipset, use those which are marketed as DUAL kit. Their supposed to have been tested and paired because of similar tolerances, the requirements for a good dual channel run.

 

However, just look at my mix..... and in 3 dimm slots at that !

Two sticks of BH-5 and one of TMD. I guess the bottom line will finally be ...DOES IT WORK ??? regardless of the mix !

[zaubara]

Hey there!

I wanted to know if that burn-in is also useful on CH-6 mems? Does the burn-in only increase the max FSB or the timings aswell? (my CH-6 are rated for 2-3-3-6, 2.5-3-3-6 on AMD systems :/ and i want to run them on 2-2-2-11 on an FSB greater 200MHz)

 

And I do need some more cooling at 3.3 volts, right?

 

Thanks!

 

edit: you might be interested in what RAMs i have Well, its a matched pair by Corsair, TwinX1024-3200C2 (rev. 4.2 i think)! Thanks again!

[cheezies]

Originally posted by zaubara

Hey there!

I wanted to know if that burn-in is also useful on CH-6 mems? Does the burn-in only increase the max FSB or the timings aswell? (my CH-6 are rated for 2-3-3-6, 2.5-3-3-6 on AMD systems :/ and i want to run them on 2-2-2-11 on an FSB greater 200MHz)

 

And I do need some more cooling at 3.3 volts, right?

 

Thanks!

 

edit: you might be interested in what RAMs i have Well, its a matched pair by Corsair, TwinX1024-3200C2 (rev. 4.2 i think)! Thanks again!

 

First off, I didn't come up with this burn-in, so hopefully someone knowledgeable about this will come and answer or questions.

 

As for cooling at 3.3V, this is a REQUIREMENT! doesn't matter whether your burning-in or doing whatever, at 3.3v, you need cooling

[uwackme]

I was doing ONE stick at a time, and doing CPC ON., 5,2,2,2.0

 

I actually used my NF7-S as the burnin mule, so the alpha timings etc were ABit default and cpc is on cause at the time there wasn't a CPC off option. But on DFI I'd do alpha's of 1,1,2,1,1,2,1 along with CPC on to be as TIGHT as possible. You dont want to give the ram cells a moments rest.

 

So 5,2,2,2,2.0; e,e,f,e; 1,1,2,1,1,2,1, auto/auto

 

Run memtest #5 and see how high you go till you hit just a few errors on a run. At this point put it in loop for 6 hours. Then up the FSB 5-10Mhz see how it is, and loop again for 6 hours. You should see upward progress of the "error" point. But 72 hours is a good ballpark for the process. This is an accelerated effort, most effective in waking up VIRGIN sticks. The last 10+Mhz FSB you will ever get out of the rams WILL take 1-2 YEARS of normal operation.

 

That is why older Ebay sticks do great and brandy new hyperX3000 sticks start out slow....the older stuff out there "used" has been burned in through use for months/years.

 

This is also why noone realized initially that BH5 was...GAWD. It took time and burnin before someone noticed HOLYCHIT BATMAN!!!

 

Cooling during this is paramount. Thats why I resorted to taking OFF the heatspredders and putting a 80mm fan directly over the DIMMs during the burnin....and for any ram "limit" testing actually.

 

Other variables enter in, what DIMM slots, what Vdd setting is best for your NB chip, what are the limitations of your CPU fsb and Vcore circuits.....when you start reaching the maximums.

 

I use DIMM1 and DIMM3, and 1.9Vdd...but your setups might like a different arrangement.

 

Also, once you have gotten the ram closer to its peak, the drivingstrength/Slewrate settings vs what DIMM slots will get you a little more FSB, its all trial and error from here.

 

 

** I didnt believe (as an engineer) in burnin on these till I saw it with my own eyes. Took 6 sticks of virgin HyperX3000 which peaked at 185Mhz and got them all to 240-245Mhz with this treatment.

 

Im looking forward to Vmod'ding my Vdimm to 3.4-3.5V so I can break the 250Mhz barrier....without waiting a year+ :nod:

[cheezies]

Very informative reply uwackme! I was waiting for your post all along! Silly thing was, I thought I had to run my RAM "hot" in order to "burn in" LOL!

 

I'll need to make a mod of some sort to mount an 80mm fan over my bh5 babies. I don't really want to take the heat spreaders off though.

[flyboi]

Shud i burn the ram with errors in it until the errors become lesser?

 

 

Cheers.

[Katch]

Originally posted by flyboi

Shud i burn the ram with errors in it until the errors become lesser?

 

 

Cheers.

 

Have you actually tried reading this post?

 

Cheers.

[uwackme]

On the errors... a few is ok, but not a bazzillion. You want the ram getting the crap kicked out of it, and too many errors, the program goes into subroutines to log them instead of hammering the ram. So a "few errors"...10-20-50-100 is ok, not not K's.

 

I like to get it to the point of say 2-4-6 errors and leave it there for 6 hours getting roasted.

 

It is good to have the ram HOT, just not so hot that you are generating errors from the interace circuits.

 

Mind you, this is ALL a hypothesis, as winbond them selves havent really come to understand WHAT was happening here...

 

But I think roasting/beating on the memory cells does some form of slight damage that actually makes them work faster/leak less. But yoou dont want to overheat the interface cicuits that drive the IOpins, you want the errors to be solely from the memory cells themselves, pushed to the limit.

 

If winbond could ever really figure out WTF was going on, they'd have a very valuable discovery. Something about the physical design of the BH5 memory cells, materials used, feature size .18um/.15um?, etc made them have this very amazing speed ability. Noone, despite SERIOUS efforts and tens of MILLION$ in R+D has figured out an equivalent.

 

If Micron or Samsung could figure it out they'd crank this . out by the tons.

 

Even with the new asskicker 939pin A64's, 250Mhz 1:1 5,2,2,2 is KILLER. Gimme 4x 512M @250Mhz quadchannel and Ill break 10K on 3D05 legit :nod:

[cheezies]

Yeah, if they could only figure out what they did, I'd buy a few more sticks myself!

 

Throughout my last 2 burn-in's, I have been using a small household fan to blow air at the side of my case. In the last 21 hours, memtest86 has reported only 64 errors. This is a lot less compared to what I got on my first burn-in. What's even better is, for the first burn-in, I ran it with looser settings too! 2-2-2-8 compared to 2-2-2-5!

 

Likewise the household fan doesn't seem to be enough, the RAM is still feels hot because air isn't directly blowing at the RAM modules.

 

*edit: that wasn't 86 errors, it was 64 errors, corrected.

*update: been running for 28 hours 46 minutes, only 64 errors, definitely much less than last time

[uwackme]

Yeah for me, I take a 80mm fan, and tie it with a tiewrap to the power cord going into the CDrom drive.....this hangs it right over the DIMMs in the case Im in. Its only temp anyway, but gets good airflow over the rams.

[cheezies]

Originally posted by uwackme

Yeah for me, I take a 80mm fan, and tie it with a tiewrap to the power cord going into the CDrom drive.....this hangs it right over the DIMMs in the case Im in. Its only temp anyway, but gets good airflow over the rams.

 

That's pretty ghetto, almost as ghetto as my household fan, LOL.

 

So far so good, it's now been 39 hours and 31 minutes, and I have 68 errors, an increase of 4 errors, ever since my last post which was recordred at 21 hours.