What is "burning in"
sociable-nomad
New Zealand - The land of the long white cloud
Ive notived everyone talkin about 'burn in software' and burning in various components... what is all this have to do with overclocking your machine?
0
Comments
F@H is a great program as it will run your CPU at 100% for as long as you want it to. 3DMark running in multiples of 15 is a good way also as it makes use of the CPU, Graphics cards and Ram... and it pushes your PSU.
But Prime 95 is the best program for O.C. testing
Correction, you don't THINK you've had any bogus WUs go to Sanford. There's no way you could possibly know if what your machine has processed and sent was perfectly correct.
Setting is dependent on the paste in question, pastes like AS5 (Arctic Silver 5) tend to increase in thermal conductivity with time. Others like Nanotherm Blue don't see nearly the benefit from setting in.
As to lowering the Vcore, I have no idea what's that about unless these are guys doing no overclocking and running silent rigs. Running lower Vcore does indeed benefit you with lower core voltages but it can lead to stability issues.
The only way to know if you've sent fubared WUs to Stanford is by early termination. If you're sending WUs that are running the entire time you can still be sending garbled data, you just won't know it.
There are really two different definitions of 'burn-in' that I am aware of.
1) The type being discussed in this thread - Running a CPU (or component) at 100% utilization for maximum thermal output. This is usually for testing purposes--how well is my cooling system working?--is my system stable? etc.
2) There is also the type of burn-in that sociable.nomad mentioned earlier. This type of burn-in is like an excercise routine for your CPU. With a new CPU, you can 'train' it to run (stable) at a lower vcore value than previously possible. This is usually done by running a CPU as close to the stability limit as possible--with as little vcore as possible. Eventually the CPU will be able to run with less vcore, or at slightly higher clock speeds. This seems to only work with brand new CPUs (less than two weeks old or so). I tried this with my new 4200+ and it did work for me.
Ideally, you should use an application such as Prime95, SP2004, OCCT or CPUBurn as they will provide error detection and are more intesive than F@H. F@H generally does a good job at detecting corrupted WUs before they are sent out but I would not use it as a stability or stress testing tool. There is really no harm in running F@H to 'burn-in' a non-OC'ed system as per the first definition. It is highly unlikely that you'll have WUs ending early in that situation.
OK ...I'm not sure what my default vcore is since it's listed at newegg as 1.30/1.35. So ...I set it to 1.325v. Then I upped my clock to 2500mhz. It seems to be stable for now but it might need to go up to 1.35v to be stable.
Anyway ...at this point do I raise the vcore if unstable or lower the clock?
I never did burn-in cause I never understood it as you can tell!
Also ...according to cpu-z my voltages fluxuate between 1.34 & 1.376 or thereabout. Overvoltage is disabled. Fluxuating vcore is very common for this board.
But, if you wish to check out my thread I did I mention it towards the end. Ideally the simple simple version is this - You teach your CPU to function on as little juice as possible it then helps with overclocking by maintaining the stock voltage when you normally have to increase, this reduces temps and puts a smile on your face
First off - BACKUP important stuff!!
run complete stock - make sure everything is stable
Then slowly decrease from the stock vcore of the CPU in as small jumps as you can make - re-run (with dual core CPU's) 2 instances of Prime95 - max heat - run for at least 10 hours - I prefer 24 to be honest.
Go down with the vcore as low as possible following this method, until you fail the in some way - test, no post or whatever.
Then go back up one step in vcore (what was stable and retest)
Then start overclocking the CPU back at stock vcore, you should hopefully see a nice increase. I managed to get to around 2650 ish with 1.35v with my Opty 170 - this is a slow process mind and requires patience.
Hope you get some success and I will hopefully finish my full guide as soon as time allows
HTH
I'try your method tonight.
So now I've got it folding 1.175v which is being reported by my monitoring software and cpu-z as 1.20v even with the overvolting disabled. I'm running at 2110 and will probably climb to 2120 tonight if all has been stable all day. We'll see.
Ok ...I'm presently at 250X10@1.325. My actual voltage under load is 1.35v average. I prime and fold for hours on end lets say 24 hours. What's my next move do I up the htt to 251 then 252 and so on til I can't get stable anymore lke lets say 255 ...then do I up the volts to 1.35v and continue til I cant go anymore lets say 260? And so on?
That a slightly different way than I advise. Does not mean its not correct though!
Why not keep it there but drop the vcore see how low you can go with 250x10.
OR
If you are wanting to crank it then sure push up the htt and see what you can get - I have found it very important to retest after EVERY bump up.
My first results from my burn-in went from having to increase stock vcore at 2.4Ghz (opteron 170) to getting stock vcore to 2.64Ghz. That took a few weeks mind, patience is required!
I am aided with my DFI mobo - its an excellent overclocker - the CFX3200 has a billion things to tweak too. Still a few things to tweak BIOS wise for rock solid stability
Hows your temperatures at the moment are you noticing any benefit??
Could I do this voltage training on my A64 X2 3800+? It currently runs at 2.4GHz, I'd like to see if I could try overclocking it again and get even higher if possible. But I have forgotten much of the knowledge I used to have.
I must have misinterpreted the whole thing. I started at stock frequency and undervolted from there. As a recap ...I volted down to 1.175v cause 1.150v gave me an undervoltage message from the bios although it would run in windows. It made me uncomfortable. So I tuck with that.
Then I began to overclock from 200x10 5mhz at a time. when I got to around 220x10 I tried going into windows but I got blue screens and freezes so I tried again at 215 and finally 210 which let me in. From there I began burn-ins with sp2004 for a few hours then I'd switch to f@h for a few more. I've let it burn anywhere from 8-12 hours maybe more but not less.
Now I'm up to 217X10 which is wonderful really ...but will it help me on the top end? LOL
I guess I'll keep on going until I can't go anymore then I'll try the method you mentioned. I can get to 250x10 @ 1.325v but it's not stable any lower than that so I'll try a burn at that setting and see what happens.
That will be in a few days of course!
RWB just try it ...it shouldn't hurt anything.
It sure does. Just about any integrated circuit can benefit--GPUs, RAM etc.
Meanwhile I upped it to 250x10@1.325 again to start there and see where it takes me. I get an average 1.35v. So what are my steps from here. Am I upping the htt or lowering the vcore? If I'm upping the htt what do I do when it won't go higher? I still haven't got this figured out.:bawling:
Nonetheless ...I have primed at 245x10@1.30v for almost 10 hours now and I will begin upping from here tonight. Let's see how it goes.
On another note ...the cpu has been running 2 weeks today. Am I sol now?