Ethernalite wrote: Why would you lap a heatsink?
Just like the idea about lapping the IHS, some ppl lap their heatsinks to get a smooth finish for better contact between the IHS and HS, and thus better heat conductivity. I consider lapping the IHS a little more risky, as you're actually dealing with a delicate component. If you lap a heatsink, then you only have to worry about the finish.
Fair enough. And Clue69Less, thanks for the correction on the whole lapping of the IHS. I've never done it personally, so I wouldn't really know about the difficulties of metal work.
Ice Czar wrote:
as a rule of thumb for every 10C you decrease the temperature you increase the life expecancy by 100% conversely for every rise of 10C life expectancy decreases by 50%
http://forums.cgsociety.org/archive/index.php/t-140765.html
Urlyin wrote:
For every 10C you increase a discrete parts (memory chips, cores, Mosfets, IC's, etc) average
operating temperature you cut it lifespan in 1/2 what every that lifespan may be. Conversely
for every 10C you lower the average operating temperature you double the parts lifespan. It
doesn't take a rocket scientist to conclude the card that runs 60C will have an average life
span that is 4 times longer than if the card runs 80C.
http://www.bleedinedge.com/forum/archive/index.php?t-5305.html
As for the original post: Don't worry about your chip suddenly dying unless you do some severe overclocking or let it get fried during a T-storm. Get a Zalman, Scythe, or Thermaltake cooler and some Arctic Silver 5. I have been using AC5 for 4 years now, and I've used it on Barton Athlon XP chips (and I heard how dangerous it could be) with NO problem. I am even using it on a A64 3000+ DTR chip, which does NOT have an IHS, and I haven't had a problem. Don't worry about the conductivity part; unless you decide to put some on the pins (and please don't), you should be fine.
EDIT: And please excuse the primitive quoting that I've done; I've never really used that feature but there are times when I really should use it, like now.
I would presume people are getting those numbers from the Arrhenius equation, which would be about in line with those numbers. The Arrhenius equation is still widely used to figure out MTBF in certain things, because it (theoretically) can be used to determine the relationship between temperature and chemical reaction rate. A chip manufacturer obviously can't wait 5 years to determine a chip's MTBF, so they will stress test it. A CMOS chip manufacturer may put their chips and run them at 125° with 100% relative humidity, determine the failure rate after a month, and then extrapolate down to what their maximum specs allow for, and determine the MTBF, and then pad that number down to what they feel is safe.
However, that being said, it doesn't really matter. Yes, increased heat will negatively affect lifespan (not quite as simple as halving the life with an extra 10°, though). Big deal. I don't think you quite respect how resiliant chips are. I don't know about AMD's chips, but Intel's design goals are for less than 1% of their chips to fail in the real world after 5 years, and less than 3% after 10. This is in real world use- where faulty power supplies exist with $23 motherboards that have a 16 cent voltage regulator that isn't running in spec. This is running in a world where sometimes fans stop working, and hell, sometimes heatsinks plain fall off. Thermal pads dry up, and people stick their computers up against the wall and have no air flow.
All I'm saying that, as long as you are running the chip within spec temperature, or at least somewhat close to it, it's not a big deal. Chances are very good that your chip will continue to chug along for another decade, even with your high temps.
Back in college, I worked part time at a local high school. I did most of the hardware repairs, and I can't think of a single time a CPU failed. Power supplies - all the time. Hard drives - even more often. Motherboards - occasionally. Processor? No, not one. I had a couple of boxes whose processor fans had failed, and the processor must have been living on the edge of thermal throttling at 90°, but the reason why the problem was noticed was because something else in the box had failed, not the processor.
I realize that ancedotal evidence is not the best way to support an argument, but all I'm trying to say is that it simply is not worth his time and money to buy some nice HSF.