Overclocking Question

[route66]

Hopefully this isn't asked a lot, but I was wondering about the term "burn in" and the effect it has on overclockability. In particular, if I buy a new processor and run it at stock speeds for a few months, does that reduce the ability to overclock the processor in the future?

 

[Furen]

Any usage will "wear out" the CPU. If you overclock you will wear it out much faster than if you run at stock. I want to say that "burning in" a cpu (that is, to run it at high voltages to make it get used to the frequency) is probably a myth, but there's some people that swear by it.

 

[Bona Fide]

I think that almost all parts need time to burn in, per se. You don't need to do anything special, just use the computer often. It's just like how a brand new pillow feels a lot harder than it's supposed to, for a few days. Then, after using it, it becomes softer and more comfortable.

 

[Furen]

^ see?

Originally posted by: Furen
[...] is probably a myth, but there's some people that swear by it.

 

[charloscarlies]

As far as I'm concerned...burn-in is a total myth.

 

[natto fire]

I thought the whole point of solid-state electronics was to avoid undesireable effects such as burn-in? (Which was probably a hit back in the vacuum tube days)

 

[RichUK]

Originally posted by: charloscarlies
As far as I'm concerned...burn-in is a total myth.

i agree, i think it is total BS, i did some research into this on the net for more information on this, and there was never any hard facts, or information that explained this theory .. some people just convince themselves that it is a needed stage which results in a higher OC .. i have tried it using some software called burn-in .. and it did absolutely nothing for me ..

it supposedly able to make the transistors switch faster, and also require less voltage to run at any given speed. .. HMM I DON?T THINK SO !!!

 

[DAPUNISHER]

Anandtech FAQs- Does "burn-in" really help overclocking?

 

[exar333]

we (overclockers) often try to describe burn-in like everyday things....like breaking in a car, or do this/that...fact is, nothing really supports it....some facts that might make ppl *think* they have burn in...

(1) thermal paste takes time to get to optimal conductivity

(2) often it takes time to figure out the optimal settings for your HW (fsb, mem speeds, ect...) so you THINK its faster, but in reality it/s really trial and error

 

[BonzaiDuck]

Well, let's scrutinize this more anal-retentively.

I am relatively new to over-clocking, in comparison to the number of year's I've been building and modding computers. I taught "Computer Architecture" and "Assembly Language" frequently over some six years during the late 1980s through the mid-1990s, but I was always a "software" person. I'm not an electrical or electronic engineer.

Personally, I've discovered that with the right choices of components, you can achieve fairly satisfactory over-clocks without changing the CPU VCORE beyond its default or "Auto" setting. So there's a range in which you are within the warranty of the processor on its voltage setting and actually right at the "stock" voltage.

Beyond that, there is a range of voltage settings that may be within the processors "warranty" range, but you also notice a slight increase in processor temperature -- maybe the idle temperature increases by 1 or 2F at a controlled room-ambient, and the load temperature also increases proportionately -- by the same amount.

It is said that heat and voltage are the primary, most significant enemies of processor longevity.

Also, I am wondering if a slight voltage increase may have a two-way influence on temperature: that is, by raising the voltage slightly, there are reasons which cause temperature to decline slightly, and reasons for which it might increase slightly. For instance, I found that load temperature was higher at the same room ambient at a lower voltage setting and the same external frequency (FSB) setting, than at a slightly higher voltage setting and the same external frequency when running PRIME95 "Large FFT, stress with heat."

So the question is this. If the warranty limit of the processor is 1.55V, if the "stock" setting seems to produce voltages at or below 1.5V, and that "idle" voltage even under "stock" settings seems a few hundredths higher in BIOS and climbing toward 1.58 or 1.6V in the OS while "load" seems to depress the value toward or below 1.5V in the OS, how much does it affect processor longevity and performance to bump up the VCORE to 1.525V if the peak values (at idle) don't significantly vary from the peak values under the stock setting?

 

[Furen]

There's no such thing as a voltage range that may be under a "warranty" range. If you intentionally overvolt your CPU or intentionally overclock your CPU (even by 1MHz) then you are voiding your warranty. If you mean a voltage at which your CPU is guaranteed to work then I'd say stock voltage +-5%, to compensate for the motherboard's inaccuracy. Though AMD and Intel MAY test their CPUs at higher voltages they will never guarantee any usage pattern that voids its warranty. If you increase your voltage you will always have more heat. No matter what. Transistors work the same way no matter what voltage they are running at and the higher electrical tension WILL lead to a higher heat dissipation. Having such a huge idle-load voltage swing indicates that either a weak motherboard or an underpowered powersupply. What truly matters is the voltage at load, since that is when a higher amount of the CPU's transistors are being utilized at their fullest.

 

[BonzaiDuck]

Thanks. First, as to the "voltage swing," it only shows under the Operating System. At stock voltage set to "auto" in the BIOS, this occurs, anyway. I've noticed the same behavior, between the same extremes of a range, for some six to ten motherboards, and an equal number of power supplies, including the OCZ PowerStream 520 and the Antec NeoPower 480. About half the motherboards are made by Intel, and the remainder by ASUS.

I have observed that one OC'er was able to push up his external frequency by 25% using water-cooling and a voltage setting of only 1.5V [manual -- not "auto"]. In my case, I had to bump up 25/1000 of a volt to get PRIME95 "Large FFT" stability. At this point, the load voltage never wanders over 1.5V.

[I hang my head in shame among AMD'ers] to note that I'm using a 3.0C Northwood (Intel). The retail box has a statement saying (I paraphrase) something like "1.55V warranty limit" or "warrantied to 1.55V" -- or maybe it says "1.55V maximum."

The difference between "load" voltages at "stock" versus that which shows after booting from a mobo setting of 1.525V -- is 0.001V -- as measured through the OS software. One would think that such measurements of differences are more accurate than any absolute reading.

However, since both the mobo setting and the measured "load" value are well below the value specified on the retail box, since the mobo setting is well outside the lower limit of your +-5% guideline, I can't see how this can result in significant shortening of processor life. Of course, I agree -- it "voids the warranty."

For temperatures, figure that other similar processors running at stock settings are used (or were used) in OEM systems with a somewhat higher case (interior) temperature and higher load temperatures.

Anyway, if the processor carton states "1.55V Maximum" (or whatever), and you set the BIOS manually to a value below that, it would seem to me that the only way one would then "violate the warranty" is to push the external frequency above its default 200 Mhz value.

 

[Diogenes2]

The fact of the matter, ( at least with P3's and Celerons ) is that Intel sold higher clocked CPU's where the only difference between Clock A and Clock B, was that Clock B had a higher Vcore.. Maybe they still do this with P4's.. I wouldn't know..

If you Bought Clock A and bumped the voltage, you now had a Clock B, even though Intel would not warrant it at that speed, of course.

Don't some of the higher clocked FX's have Higher Vcore than some of their slower versions ( same fab, of course ) ?


Intel used to write on Celeron Boxes that they were not for Dual applications, even though they worked just fine in the BP6..


If a CPU is working, it is not OVERclocked. Once it's OVERclocked, it stops working.

If my 4200+ is running 2.5g at 1.35v it is not overclocked. It's just a 4600+Ultra , that had the multi set to 11 and put in a box labeled " 4200+ "

 

[Furen]

BonzaiDuck, sorry about that, I was assuming that you were talking about an A64 since the voltages quite close to that and prescotts are way lower. The 1.55v thing on old P4s is basically the max voltage that will be used within a family or series, I think. Like I said though, overclocking voids your warranty, so unless you just want to overvolt for the hell of it it doesnt really help. That said, manufacturers have no way of knowing whether you overvolted or overclocked a CPU so as long as you dont admit it...

 

[BonzaiDuck]

Oh, sure, I understand that.

I'm just speculating that there are a several ranges of OC'ing options, and one of those should not -- theoretically -- shorten the life of the processor.

You'd only need to exercise a lawyer's honesty D ) if you do, in fact, cause the processor to blow before its three-years (or OEM one-year) are over.

 

[RichUK]

OEM is only three months i think ^^^^

 

[Kogan]

I've been using and overclocing computer components for about 10 years now. After about a week of "burn-in" (running prime95/memtest/etc), typically, most of the components I've used did not change - they worked just as good as on the first day as they did a year later.

Except for these instances:

AMD XP 2400+ cpu. When I first got it, it was only overnight prime95 stable overclocked at about 1.7v. After weeks of use, it was all of a sudden stable at 1.6v at the same overclocked speed.

AMD 64 venice cpu. When I first got it, it was only 100% stable at 2700mhz 1.55v. After weeks of use, it was stable at the same speeds at 1.475v.

TCCD ram - got this brand new ram hoping for ddr600(300mhz) (all testing done at 2.8v). Was dissapointed that it would only boot into windows as high as 270mhz (unstable).

For day 1, I have records of the ram giving an average of 136 errors in memtest test #5 at 270mhz 1t 2.5-4-4-10).

After a few days of heavy use at stock speeds/voltages, I tried again:
For day 3, It gave an average of 8 errors in memtest #5 at 270mhz (1t 2.5-4-4-10).

Several weeks later, the ram runs flawlessly at 270mhz+, and even boots into windows as high as 310mhz.


Just my experience Maybe only brand new hardware that hasn't been tested extensively by the manufacturer does this. And burning-in to me means just a few days/weeks of normal use and prime95/etc - not excessive volts or running the component at un-stable speeds.

If overclocking, don't worry about burning-in the component. Just find your max stable overclock and stick with it. If you want to see if burning-in helped, try a higher speed or lower voltage after a couple of weeks and see what you get.

 

[evetstech]

For CPU burning in, "burning in" essentially makes the thermal paste in between the IHS and the core sort of go on better. It's just like when you replace AS5, you need to essentally "burn in" the AS5, before you see noticable differences.

For Ram, I still have no idea how that works. It works for some, it doesn't for others. I think TCCD, burning in will generally not help at all, but voltage liking ram like BH-5 usually works