CPU burn in

[NumericalMethods]

Hey guys,

I've been a lurker for a while and now I actually have a question that I haven't found too much info on.

CPU burn in. Often mentioned but never really discussed in technical detail. As far as my limited understanding goes, during an initial bed in period, a CPU will 'get used' to running at a specific frequency. That could mean that overclocking a chip that has been running on stock speeds for a long time becomes more difficult.

Is that correct? What actually happens to the gates on the die that tunes a chip to a specific frequency? How long does that process take? - Hours/ days/ months.

I ask because I'm about to install a q9550 and this will be the first chip that I have overclocked from box fresh (after checking that it actually works). Just wondering what would be going on under the IHS as I slowly turn up the heat!

 

[z1ggy]

All I can really offer you is that when I burn burn my cpu in when I OC, I do the fsb increase before I go to bed, and then when I get back from work tthat next day ( about 18 hours later) I run the stress test. If it passes..then I increase the fsb more..repeat.

 

[Idontcare]

Originally posted by: NumericalMethods
CPU burn in. Often mentioned but never really discussed in technical detail. As far as my limited understanding goes, during an initial bed in period, a CPU will 'get used' to running at a specific frequency. That could mean that overclocking a chip that has been running on stock speeds for a long time becomes more difficult.

Is that correct? What actually happens to the gates on the die that tunes a chip to a specific frequency? How long does that process take? - Hours/ days/ months.

You have correctly identified what people seem to envision cpu burn-in to entail.

But it is a false view of what is going on.

CPU's do not "settle in" to work at a specific frequency. The only product in this industry that requires burn-in for performance reasons is Artic Silver 5.

Burn-in time in electronics is solely related identifying early fails. Checkout this anandtech review page regarding warranties and the like. Burn-in removes the lefthand side of the distribution of early fails by forcing them to fail before being put into mainstream usage.

Burn-in was popular/common back in 386/486 days as Q&A methods at the manufacturers was still a work under development at that time. By the pentium era the quality at Intel/AMD was good enough to catch the majority of early fail suspects so the penalty for not doing burn-in at the consumer level was dramatically reduced and it became a "feature" sold by the DELL's and likes as a premium service to have your rig burned-in prior to shipping.

Nowadays it is really just done by DIY'ers who want to make sure the parts they bought are not going to fail right away or perform improperly before they get too excited about OC'ing.

But as far as the concept that your CPU needs to "settle" into a groove to operate a certain frequency, pretty much complete BS. It's one of those ideas that I agree sounds plausible and reasonable to a lay person, but once you know some things or two about IC's you come to realize it's just not a true phenomenon, there is no basis in the physics of its operation for the concept of physical/electrical burn-in as the term is envisioned here.

 

[Gillbot]

I never bought the hype of burn in, I crank my chip to the max as soon as I take it out of the package!

 

[NumericalMethods]

Thanks Idontcare. I was somewhat suspicious that there would be a physical change in the chip, so I was really struggling to see how the phenomenon would occur. That makes sense about early to fail chips.

@ Gillbot - lol I have seen you mention this before! I assure you I will be using stock settings only for a fresh install of my OS and to get an idea of what temps the chip gives me at stock... after that, who knows? Here's hoping I get a good one

 

[Idontcare]

Originally posted by: NumericalMethods
Thanks Idontcare. I was somewhat suspicious that there would be a physical change in the chip, so I was really struggling to see how the phenomenon would occur. That makes sense about early to fail chips.

Not that the info changes the reality of our burn-in discussion, but I feel I must state for completeness that down in the minutia of the nitty-gritty there is technically a physical change in the processor in terms of the location and concentration of defects (impurities, voids, charge traps, etc) that does occur during any usage of the device and burn-in (true burn-in, not the misconception type we are discussing here) is intended to accelerate this to force early fails to fail early.

Again this reality (subtle physical changes at the atomic level) is not to be conflated with the misconception that an integrated circuit requires a "burn-in" process in order to optimally or efficiently operate at any given clockspeed.

 

[NumericalMethods]

Ok cool. Completeness is important. I guess in my imagination, somewhat limited by lack on knowledge, I was imagining the substrate 'thinning' between the terminals causing the switching properties of individual transistors to change. This is where I thought things didn't really add up. Thanks again.

 

[Gillbot]

Originally posted by: Idontcare

Originally posted by: NumericalMethods
Thanks Idontcare. I was somewhat suspicious that there would be a physical change in the chip, so I was really struggling to see how the phenomenon would occur. That makes sense about early to fail chips.

Not that the info changes the reality of our burn-in discussion, but I feel I must state for completeness that down in the minutia of the nitty-gritty there is technically a physical change in the processor in terms of the location and concentration of defects (impurities, voids, charge traps, etc) that does occur during any usage of the device and burn-in (true burn-in, not the misconception type we are discussing here) is intended to accelerate this to force early fails to fail early.

Again this reality (subtle physical changes at the atomic level) is not to be conflated with the misconception that an integrated circuit requires a "burn-in" process in order to optimally or efficiently operate at any given clockspeed.

Sorry, I just thought about woodchucks chucking wood when I read that! :shocked: :laugh:

 

[VirtualLarry]

Originally posted by: NumericalMethods
CPU burn in. Often mentioned but never really discussed in technical detail. As far as my limited understanding goes, during an initial bed in period, a CPU will 'get used' to running at a specific frequency. That could mean that overclocking a chip that has been running on stock speeds for a long time becomes more difficult.

Nope, no such thing.