How can Intel Know Who's been naughty or nice?

liveonc

Distinguished
Mar 24, 2008
437
0
18,780
Hi all! No wise words, for I know not the answers. Just a question I'm dying to have answered.

How can Intel, or for that sake AMD, know who's OCéd their CPU? There's a 3 year warranty on those Intel chips. So okay, you stay within the specs, you don't open, scratch, or throw the thing arround. But you OCéd and the thing dies within 3 years. What then?

Or how about? You OCéd over/under volt the thing, and it got hot but didn't char. How can they tell? Can they? If so how???

Need answers here. Even insiders are welcome to enlight!
 

Evilonigiri

Splendid
Jun 8, 2007
4,381
0
22,780
Great question. I have only an idea.

Let's say you went beyond intel spec, letting the CPU hit 90C and raising the voltage to 1.8V. This cpu will die within 2 years with normal use, I'm quite sure of it. The cause of the death would be electro-migration, in which the high voltage and heat would speed the process up. I'm sure Intel will check for signs that the cpu has died from electro-migration (from the wires and such) and conclude that the cpu was not within Intel specs.

I'm also quite sure that more than 99% of the cpus (in retail) is not defective, so the chances of a cpu just dying from defective parts is slim and none.

Of course there are people who uses their pc day and night, stressing the components continuously, and perhaps a weaker part of the cpu would give away. I'm not sure it would die from electro-migration though.
 

liveonc

Distinguished
Mar 24, 2008
437
0
18,780
can electro-migration only be sped up through high temps and voltage? What about ambitious FSB or 24/7 use? I like using pc to do folding if I'm not using it. How does under-volting effect electro-migration? My first pc was an 86 from 86. It was 24/7 until the 286, but was still used. Last I checked, the thing still lives:)
 

Evilonigiri

Splendid
Jun 8, 2007
4,381
0
22,780
86 huh? Well, the manufacturing process is much larger, so electro migration isn't a large issue for it. As the manufacturing process gets smaller and smaller, electro-migration becomes a much larger issue. 45nm is where we have to start seriously worrying about it. As we get to 32nm...we'll see.

Every electrical component suffers from electro migration. Heat and voltage is the key to reducing the process, so under-volting will slow electro-migration down (Less voltage = less current = less particles being 'migrated'). Lower voltage also lower heat, so the effect is exponential (my guess).
 

liveonc

Distinguished
Mar 24, 2008
437
0
18,780
I've read (without digesting), before about electro-migration. But all the previous mention of this gave all the blame to a higher vCore. The article you linked to explained that, in contadiction to the articles I've read before, where the great satan was voltage, that higher voltage actually can prolong the life (as long as the increase in heat due to the higher voltage was properly cooled). I've been buying new pc's anually since 2003, but bought laptops 2 years in a row, and before that couldn't be bothered to OC. So it's first now that I'm trying to figure out how. Even though I only need my pc to live a year, I've given my former pc's to relatives who use it until nothing will run on it anymore. With me it's always been the mobo that died, but I've always been cheap on the mobo's. I think that the P5K that I just ordered, "on special", is the most expensive mobo I've ever purchaced if you compair to what prices are on average during the different years. The most pricey was my P1 and PII, but back then it was cheap.