overclokcing, electromigration, urban legend

[ensabah6]

hello,
is it true that an overclocked cpu, over the months and years, due to "electromigration" becomes undergoes a physical change so that it more easily accommodates the overclock? or is this an urban legend.

i've seen this claim on this forum, but is there any basis in engineering and science for this?

dan

[SK8]

electro migration kills cpus so in other words it doesnt accomodate the overclock. It undergoes a physical change that shorts out micro paths in the cpu which in effect kill the cpu, there is a problem with northwoods being fed over 1.7v , SNDS.

[aNTiChRisT]

I believe so, but im afraid i have no proof. My old 2.8 would do 3.1~ but after about a month reached 3.2 stable on the same volts. Go figure :P No i didnt change anything, but no proof either.

~t0m

[Enigma422]

SK8 is right in saying that electromigration kills your cpu. What electromigration is is the redistribution of the atoms that make up the interconnect in your chip due to high electric field and current density. Over time this could lead to voids in the interconnects creating open circuits in your wiring (open circuit in wiring = very bad).

In fact one of the major reasons the semiconductor industry has moved away from aluminum interconnects to copper interconnects is that copper has higher resistance to electromigration.

[batboy]

You are confusing "burn in" with "electromigration".

High voltage and heat to a lesser degree, will accelerate electromigration (which is very bad).

[ensabah6]

the reason i posted this thread, is that i've heard of responses like MikeC, and i wonder if it is not electromigration, what is causing this phenomenon

mike c quote

"I dunno about that. I have 3 stories:

1) A P3-550 (slot FF) -- oc'd fairly well to ~700 but unstable beyond. Undervolted by not more than 0.15V. Kept playing with it, kept it oc'd for ~8 months. By then it did fine at 800MHz and undervolted better as well. Now running at 733 @ -.2V undervolt perfectly stable (in someone else's hands). Same motherboard and RAM throughout.

2) Celeron 400 (slot) -- very similar as above %-wise.

3) T-bird 1G -- the CPU that drove me to silencing. It would not even hit 1.15G for the longest time. After a year, it did 1.3G fine at .1V above stock. Ran it after that at 800MHz and -.25V undervolt for a long time. Finally cracked the core some months ago - HS w/bad clip.

I have read others' experience with this at overclockers.com back in the dark side days."

[medo145]

i may have been a victem of this phenomenon.

i've had my 2.4 xeons oced to around 3.35ghz for a few years now, and recently it started crashing. I had to turn down the oc to about 3.1ghz and im not even sure it's stable at these speeds. Looks like I should start looking for new chips. :-/