Burn-in works!

[robertk2012]

Originally posted by: pm
I've worked for Intel for 11 years, before that I interned with Inmos Corp. Mostly I've done design, but more recently I have been working in "back-end" debug for several years. Among other things recently, I've been writing tests for the burn-in ovens at Intel and looking at burn-in fall-out (why parts on a product are failing in burn-in) and analyzing large volume burn-in data (literally hundreds of thousands of parts). This is microprocessor burn-in - high temperature, high voltage testing done to try to eliminate "infant mortality. I am about as close an expert on this subject as you are likely to find posting at Anandtech and I'm the author of the FAQ entry that was quoted above.

Note that I am not speaking on behalf of Intel - I am not an Intel company representative - nor is anything that I'm writing specific to Intel CPU's... all CMOS circuitry from DSP's to ASIC's to CPU's all are like this.

I do not dispute anything in this thread. But what I will say is that this burn-in effect is not affecting anything on the CPU due. The silicon is not speeding up. Whatever effect you are seeing, it has nothing to do with the microprocessor silicon speeding up. In burn-in on CMOS circuitry (high-voltage, high-temperature operation), the transistors will always slow down by a small but measureable percentage. NMOSFETs slow down due to impact ionization creating trapped charges in the oxide of the gate shifting the Vt upwards. PMOSFETS slow down to PMOS BTI creating positive charge interface traps at the oxide-channel boundary which also shifts the Vt and reduces saturation current.

Beyond that, if this technique for burning in a part seems to help, then my only input is that it's not the silicon speeding up. I don't see the technique as posted by the OP as doing any harm. If it helps, then it's a good thing.

I dont see it doing harm to the cpu but im not a big fan of running a computer in an unstable state. To many things that can go wrong. If you do it I would just make sure there isnt any data that is important on the computer or make a backup of said data. 99.9% nothing will happen but I wouldnt want to be that .1%. I cant see it doing any harm to the cpu unless you are running it a high voltages which is not what is going here or at least shouldnt be.

 

[pm]

Originally posted by: robertk2012
I dont see it doing harm to the cpu but im not a big fan of running a computer in an unstable state. To many things that can go wrong. If you do it I would just make sure there isnt any data that is important on the computer or make a backup of said data. 99.9% nothing will happen but I wouldnt want to be that .1%. I cant see it doing any harm to the cpu unless you are running it a high voltages which is not what is going here or at least shouldnt be.

I agree. In between my original posting and yours I editted my post to mention the possibility of data corruption. I agree with you... if you have data that you value, running for long periods with the CPU in an unstable state you never know what kind of junk will get written back to the hard disk which can lead to data corruption of files, the OS, or the entire hard disk.

 

[robertk2012]

Originally posted by: pm

Originally posted by: robertk2012
I dont see it doing harm to the cpu but im not a big fan of running a computer in an unstable state. To many things that can go wrong. If you do it I would just make sure there isnt any data that is important on the computer or make a backup of said data. 99.9% nothing will happen but I wouldnt want to be that .1%. I cant see it doing any harm to the cpu unless you are running it a high voltages which is not what is going here or at least shouldnt be.

I agree. In between my original posting and yours I editted my post to mention the possibility of data corruption. I agree with you... if you have data that you value, running for long periods with the CPU in an unstable state you never know what kind of junk will get written back to the hard disk which can lead to data corruption of files, the OS, or the entire hard disk.

 

[allies]

After trying to run my CPU at 2.7 GHz with 1.425 v (after 24 hr burn in) I couldn't. Chillen at 1.415 v at 2.6 GHz isn't half bad, and this is what I was at prior to burn in.

 

[firewolfsm]

well, all i can say now is that i hope it works, with the voltage at 1.5 instead of 1.55, the load is 39C load instead of 42C, i'm at two hours of burnin, i'll trty prime after 6.

 

[robertk2012]

Originally posted by: allies
After trying to run my CPU at 2.7 GHz with 1.425 v (after 24 hr burn in) I couldn't. Chillen at 1.415 v at 2.6 GHz isn't half bad, and this is what I was at prior to burn in.

Your saying it didnt help you right?

 

[firewolfsm]

i don't know or care why, but now it succesfully ran at 1.5V for 5 minutes, not really stable, but it only ran for a few seconds before. that was after a 5 hour burn in on my 144 at 1.55 (1.5) V and 2.68GHz. i'm going to do 5 more hours tomorrow.
thanks

edit: i think the stock amd paste did thicken because i'm loading at 38C again

 

[Jeff7181]

LOL... not this again... come on guys... there's been facts supplied by people in the know right here on the forums that explain why "burn-in" is not something that actually occurs. The only thing this is doing is breaking in the thermal compound.

 

[robertk2012]

Originally posted by: Jeff7181
LOL... not this again... come on guys... there's been facts supplied by people in the know right here on the forums that explain why "burn-in" is not something that actually occurs. The only thing this is doing is breaking in the thermal compound.

LOL thats whay I have been trying to say. Its like talking to a brick wall.

 

[Jeff7181]

Originally posted by: robertk2012

Originally posted by: Jeff7181
LOL... not this again... come on guys... there's been facts supplied by people in the know right here on the forums that explain why "burn-in" is not something that actually occurs. The only thing this is doing is breaking in the thermal compound.

LOL thats whay I have been trying to say. Its like talking to a brick wall.

It's ridiculous... I don't understand why people think things are any different now than they were 3 months ago the last time there was a thread about this on the forums. The notion that you can somehow make your CPU run faster or cooler by forcing it to run in unstable conditions for extended periods of time is pure nonsense. People throw around the term "burn-in" way too loosely. When using it to describe what happens when your thermal paste is breaking in and your temps are gradually decreasing, ok... that happens... it's a fact. But if you're using it to say that you're burning the path electrons take through the CPU to make them go faster... that's just idiocy.

 

[Vegitto]

Oh, the setting in of the thermal interface? I think not. I was challenged to try again when I reseated the HSF and re-applied the AS5, and guess what? No change at all in temperatures.

 

[buzzsaw13]

Originally posted by: Jeff7181

Originally posted by: robertk2012

Originally posted by: Jeff7181
LOL... not this again... come on guys... there's been facts supplied by people in the know right here on the forums that explain why "burn-in" is not something that actually occurs. The only thing this is doing is breaking in the thermal compound.

LOL thats whay I have been trying to say. Its like talking to a brick wall.

It's ridiculous... I don't understand why people think things are any different now than they were 3 months ago the last time there was a thread about this on the forums. The notion that you can somehow make your CPU run faster or cooler by forcing it to run in unstable conditions for extended periods of time is pure nonsense. People throw around the term "burn-in" way too loosely. When using it to describe what happens when your thermal paste is breaking in and your temps are gradually decreasing, ok... that happens... it's a fact. But if you're using it to say that you're burning the path electrons take through the CPU to make them go faster... that's just idiocy.

Stupid or not, it works for him and he's got a screeny to prove it.

 

[DrMrLordX]

The problem is that it doesn't work for very many. Nobody seems to be able to figure out why it works for people when it does work beyond the TIM-curing explanation.

It didn't work for me, that's for sure.

 

[Jeff7181]

Originally posted by: Vegitto
Oh, the setting in of the thermal interface? I think not. I was challenged to try again when I reseated the HSF and re-applied the AS5, and guess what? No change at all in temperatures.

Did you clean every microscopic bit of thermal compound out of every microscopic pit in the metal heatsink and metal IHS? I think not. Arctic Silver themselves tells you there's a break in period... if you're not experiencing it, you're applying it improperly.

*EDIT* EVERY time I've removed and installed a heatsink, the thermal compound has taken a week or so to work it's way into the little air gaps. If you didn't experience the same thing, you either did a piss poor job cleaning the surfaces, have always installed it improperly in the past and still are, or you're lying to make you fantasy a reality.

 

[LiquidImpulse]

Originally posted by: StrangerGuy
"Step 2: Reboot at a voltage at which Prime95/StressPrime2004/Stability testing tool of your choice fails within seconds. Literally seconds."

Chances are, if Prime95 fails within seconds, Windows will most likely crash before you get the chance to set affinities.

EXACTLY what happened to me. ran for literally 30 seconds, then windows froze up, had to boot up, gave up burn in.

 

[buzzsaw13]

Holy crap! It actually worked! I set it at 1.4V@2.8ghz and let it burn in overnight, then ran SP2004 with small FTTs for 40+ hours and it was stable!

My temps dropped from 36C idle 52 load to 32C idle 46C load, I will be bumping up the voltage back up to 1.45V to see if I can squeeze more than 2.8ghz out of it

 

[robertk2012]

Did you try to see if it was stable before you ran the burn in or had it been awhile since you had tested stability?

 

[buzzsaw13]

Originally posted by: robertk2012
Did you try to see if it was stable before you ran the burn in or had it been awhile since you had tested stability?

Was prime stable at 1.45V, ran SP2004 on small FTTs for 43 hours, then ran SP2004 again after the burn in for another 40+ hours. I ran prime not too long ago as I havent had this chip for that long.

Tried the burn in with 1.35V but no go on that

 

[robertk2012]

So the last time you had ran prime was when you first installed the chip? About how long have you had the chip? What kind of thermal Paste?

 

[buzzsaw13]

I've had the chip since early Feb. I've been using AS5, but IHS is removed if that means anything and I've switched out heatsinks twice since I've had the chip.

 

[Duvie]

Originally posted by: Vegitto

Originally posted by: DAPUNISHER
Since Andy isn't here anymore, I'll do the honors You've been FAQed!

It's not the same. I'm not giving it more volts, I'm giving it less volts.

True!!! It is like starving it and lowering its metabolic rate...Eventually it learns to deal with less...Scientifically still sounds goofy when translated to cpus...but oh well...

I think Lucifer helped him!!! Now it is a grand conspiracy of deceit!!!

 

[Duvie]

People also remember in the past I have seen this this worked better with new chips which hadn't been overvolted or OC'd for an extended period of time already....

 

[robertk2012]

Originally posted by: buzzsaw13
I've had the chip since early Feb. I've been using AS5, but IHS is removed if that means anything and I've switched out heatsinks twice since I've had the chip.

When was the last time you switched out heatsinks? Artic silver takes awhile to cure so I imagine thats what your seeing happen. You might even get more performance out of it over time.

 

[LiquidImpulse]

btw, after burning in for 24 hours. it did nothing to drop my temps

 

[Maximilian]

I really really think the explanations for people seeing an increase in speed should be stickied. Same with that FAQ I was gonna try it and then read the FAQ that said it would harm my chip. Probably shorten its life.

It would stop people shortening the lifespan of their chips.