Quote:
Originally Posted by
AMD4ME 
The voltage/current is what causes metal migration. It's an electrical process not a thermal process. If you run the temp any where close to it's designed operating temp. it will run ten people lifetimes before you'd notice any metal migration.
This is not correct.
Temperature
does have an influence on electromigration just as it has an influence on resistance.
http://www.csl.mete.metu.edu.tr/Electromigration/emig.htm
http://en.wikipedia.org/wiki/Electromigration#Thermal_effects
Temperature has a significant effect on electromigration at any temp range higher than that at which the metal becomes a superconductor; unless you are running very near absolute zero, temperature is influencing electromigration in copper.
Quote:
Originally Posted by
AMD4ME
Yes if you were trying to run your CPU @ 200C then temp. might be a valid issue in regards to accelerating metal migration but there is a problem... CPUs generally shut down long before any thermal damage is done let alone metal migration.
IME it's simply not an issue when stress testing.
When was the last time you heard of someone's CPU failing during stress testing from running at the 24/7 approved temp. for the CPU? I've never heard of it and never experienced it in 20+ years of building and testing highend PCs. You're worrying about a non-issue.
Most CPU failures are caused by electromigration and electromigration is happening, on some level, almost anytime the part is powered on.
I have had CPUs fail at low temps, while operating within specificed clocks and voltages. It's rare, but it does happen.
It happens much more quickly at high temps, high current loads, and high voltage, but can still take years.
He's not worrying about a non-issue. Higher temps lead to failure sooner, and stress testing is most certainly a cause of wear and tear. It's still necessary if you want to have
Indeed every 10C increase in temp roughly halves component lifetime. A CPU running at 50C is likely to last nearly eight times as long as one running at 80C.
http://en.wikipedia.org/wiki/Arrhenius_equation
http://en.wikipedia.org/wiki/Black's_equationQuote:
Originally Posted by
UndyingEcho
Well Stress testing in cool in environments is not gonna make your overclock any less stable it will just force you to keep a hawk's eye on temp's when your environment returns to normal.
Example: Let's say its a hotter than average day and my CPU is just idling it is going to be idling a little hot but nonetheless still within an acceptable temperature range. Let us also say on said day I fire up Starcraft 2 In the settings it lists which settings are GPU extensive and which one are CPU extensive instead of trying to underclock back to a better temperature just drop the CPU extensive settings all the way down your game might not look as sexy but it is way faster than trying to re clock your system, and if your temps do hit to high of a temperature well then it's a nice day go out and ride a bike or something.
There is no hard line between stability and instability because instability is probabilistic. The point of stress testing isn't to make you feel better about a hastily done, marginally stable OC, it's to put the hardware through it's paces, executing as many instructions as possible in the shortest period of time possible, in situations where instability is most likely to be revealed. Why? Because if there is any potential for instability at all, you are testing every time an instruction is executed, every time a transistor switches, no matter what you are doing. You want errors to occur as little as possible, and if they are going to occur you want to force them to occur during your test so that you can detect them and fix whatever is causing them.
Also, just because I'm not at my computer doesn't mean it's not working. A particularly long encode job can take days, both my CPU and GPU have worked on distributed computing programs for months on end with no interruptions, often while I was away doing something and not in a position to babysit the system.
Also, heat doesn't imply a nice day. I'm currently in the south and triple digit F ambients with high 90s% humidity are not uncommon.
Edited by Blameless - 4/2/12 at 3:48pm
