localhostrulez wrote:What do you mean your 4690K died? How?
I keep wondering, since my mom does need a new laptop in a few months - is it safe to buy a Skylake chip (it'll be run stock, microcode updates, etc), or should I go out of the way and find a Haswell/Broadwell machine?
Dead 4690k - it started burning way too much power, and that was strongly correllated with a wave of glitchiness. I still haven't figured out how to entirely fix my fonts.
My concern here shouldn't apply to laptop chips (unless the thermals are nuts), since they run at much lower voltages.
CScottG wrote:Remember: Haswell had an integrated voltage regulator, but Skylake doesn't (..and Kaby Lake, at least at a 14mm process, probably won't either).
-it's likely that an integrated voltage regulator is more likely to fail when "pushed" beyond it's voltage limits DESPITE good thermal performance (from excellent cooling).
For Skylake then it should be more about thermals, and its 14mm process along with perhaps its non-use of the integrated graphics aid better thermal performance. Then, so as long as you keep it cool and stable, it should last a LONG time (..and if something is going to "give-out" it would more likely be the m-board).
FIVR MOSFETs degrade according to the same general rules as the rest of the processor. Each increase of 10C makes things wear roughly twice as fast. I'm still trying to figure out the nuances of voltage wear acceleration, but it's clear that there is still some kind of more general rule (that is, the biggest per-device variable by far is just maximum wear capacity). Also, whatever the rule for voltage is, it's pretty impressively steep. Some data I was just looking at for bigger devices showed 4.0V causing 5000x the wear of 3.3V.
Also, (assuming it's a typical enough buck converter,) the voltage more relevant to FIVR wear is what it gets from the mobo, not what it's giving to the cores. It does also depend on what parts of the chip are active at the time, but most of the stress on the chip will be from when the chip is fully active anyway. The only case that comes to mind in which this would actually matter is if the chip is used mainly for heavy but single-threaded tasks, in which case individual cores would accumulate wear much quicker than the FIVR (because of localized thermals if nothing else).
biffzinker wrote:Still have the FIVR set to the default 1.8v even with the overclock I've been pushing (4.9Ghz/1.372v.)
The datasheet max for FIVR input is 1.86V, FWIW. (Stock is 1.75 though, so 1.8-1.85 should be fine.)