12:06 <SpeedEvil> Is anyone aware of any research into how long ICs could live?

12:07 <SpeedEvil> That is - is a 500 year operating life implausible? 5000 year? Does this vary by geometry or what?

13:46 <_Sync_> SpeedEvil: chip life mostly is decided by the gates degrading

13:46 <_Sync_> at least in high power computing

13:46 <_Sync_> memory will slowly loose charge through leakage

13:47 <_Sync_> and currently you look more at 4 years than 500

14:03 <SpeedEvil> I don't mean flash memory losing charge

14:04 <SpeedEvil> Or high power stuff.

14:04 <SpeedEvil> I have many operating computers over 4 years old.

14:04 <SpeedEvil> My oldest is ~20

14:08 <_Sync_> sure

14:08 <_Sync_> but that is the current target for lifetime

14:08 <_Sync_> track your Vcpu

14:08 <_Sync_> and you will see the degradation over time

14:09 <_Sync_> well, what else could you mean? old logic crap? that's going to last fairly long

14:09 <_Sync_> and everything else has memory of some sort

14:24 <SpeedEvil> A general question.

14:25 <SpeedEvil> If I'm willing to reflash flash memory, how long will existing commodity stuff that is not high power last. Neglecting flash, how long might operating electronics last.

14:30 <whitequark> capacitors

14:36 <SpeedEvil> Electrolytic capacitors do have issues.

14:36 <SpeedEvil> There are alternatives

14:36 <whitequark> anything mechanical

14:36 <whitequark> varistors

14:36 <whitequark> anything that degrades under UV

14:37 <whitequark> any plastic from which plastifier effuses

14:37 <whitequark> if we're talking about MIL-SPEC hw made from ceramic and bakelite, probably quite long

14:37 <SpeedEvil> All of that can be worked around. If there is some fundamental limit in the silicon that I'm missing, that can't.

14:38 <whitequark> well, you have the arrhenius law

14:45 <SpeedEvil> Hmm.

14:46 <SpeedEvil> I guess I could just punt, and cheat.

14:47 <SpeedEvil> The honeywell SOI processors and other stuff that are designed for 250C operation

14:51 <SpeedEvil> hthttp://www51.honeywell.com/aero/common/documents/myaerospacecatalog-documents/Missiles-Munitions/HT6256-High_Temperature_32K_x_8_Static_RAM.pdf

15:27 <_Sync_> well SpeedEvil

15:27 <_Sync_> it's a hard question

15:27 <_Sync_> in the end it will be limited by the diffusion of the dopants in the Si

15:27 <_Sync_> most probably

15:28 <SpeedEvil> I guess.

15:29 <_Sync_> another thing will be electromigration in the backend

15:30 <SpeedEvil> Isn't electromigration strongly affected by temperature too?

15:30 <SpeedEvil> hmm -

15:31 <SpeedEvil> I guess it's to some degree not so much of an issue on an 8031@50mA rather than an i7 at 90W

15:31 <_Sync_> well

15:31 <_Sync_> you cannot compare the two

15:32 <_Sync_> one has a average heat flux higher than the surface of the sun

15:32 <_Sync_> the other one does not

15:32 <_Sync_> also one is highly integrated (by 2015 measures), the other one is not

15:32 <_Sync_> the killer is the current density

15:33 <SpeedEvil> yeah.

15:34 <SpeedEvil> It would be interesting to actually autopsy - say - pentiums that have been on for 20 years

15:34 <SpeedEvil> But that would require a lot of effort, and really needs a lot of knowledge as to where the wear might occur

15:34 <SpeedEvil> hmm - someone has to have done this

15:35 <_Sync_> the effort is pretty low, but you won't find much electromigration

15:35 <_Sync_> because it can be effectively suppressed

15:35 <_Sync_> as I said, overclock your i7 for a while and watch Vcpu

15:35 <_Sync_> that's the easiest thing to see degradation

15:35 <SpeedEvil> I've never actually overclocked a CPU since my 486 :)

22:08 <berndj> SpeedEvil, do you want to build an interstellar probe that wakes up when it reaches another star system?

22:09 <SpeedEvil> I have no specific application.