5 ms·

Electrolytic caps (especially aluminum) do generally have a pretty bad lifetime. Their specs (capacitance, and internal resistance aka ESR) degrade significantl

Electrolytic caps (especially aluminum) do generally have a pretty bad lifetime. Their specs (capacitance, and internal resistance aka ESR) degrade significantly over time, non-linearly, as temperature (and voltage) go up. At their rated voltage/temperature, lifetime maybe be ten thousand (or less) hours...barely enough to last a full year of 24x7 operation. This is one reason mining GPUs can be a questionable buy. The miner may have operated them at "low" temperature, but they were on 24x7, whereas a gaming PC may be only on a couple of hours a day a few times a week. Ultimately it comes down to how much the designers overspec'd components and/or factored in for wiggle room and how much the capacitors can degrade before circuits start acting up enough for the user to notice.

TedDoesntTalk 4y ago
> a gaming PC may be only on a couple of hours a day a few times a week Never met such a gamer
- zabzonk 4y ago
  a gamer with a career and a life outside of "gaming"?
  - Dylan16807 4y ago
    Doesn't mean you turn anything off.
- coryrc 4y ago
  More seriously though, lifetime of a component is generally halved for each ten Kelvin increase. So idling or lightly taxing the card at, say, 60°C instead of 90°C is a 8x longer lifetime alone.
anyfoo 4y ago
I corrected my text above to read "unreasonably short lifetime" now. In my experience, plenty of products from the 90s and older still seem to work absolutely fine despite their use of electrolytic caps. Yes, some of their parameters may have degraded, but it still seems to be within tolerance for their particular use in the design. It also depends on how exactly the caps are used. Caps close to a massive heat source, like a power resistor, are likely to dry out relatively quickly, while others, like those used as part of small signal filters, may last literal decades, with not much change in e.g. ESR or capacitance (and hence still conform to factory calibration pretty okay). Admittedly I don't have any hard data for that, only anecdotical experience with measuring ESR and capacitance and saying "this looks fine". The crucial point is that the "capacitor plague" affected much more than products which were on 24/7 or which had tight tolerances, to the point where it was very well visible to the "casual" consumer. Yes, we may not expect a server mainboard that was on for the last 20 years to still have its caps within spec, but wrt the capacitor plague, the gaming PC you mentioned would have been quickly affected, in an obvious manner, as well.
- coryrc 4y ago
  > while others, like those used as part of small signal filter I wouldn't expect those to be electrolytic, though. Electrolytic capacitors are mostly used when the exact characteristics are not important, like on a supply bus, where your qualifications are typically either "at least" or "no more" than some value. Tripling capacitance usually has no negative effect on a supply bus but huge change of doing signal conditioning, so something like that is more often going to be a capacitor that is intrinsically more stable.
  - anyfoo 4y ago
    Good point, but in audio applications for example, for cross-over filters, bipolar electrolytic caps are common. The “calibration”, if any, may be very simple, though.
kqr 4y ago
> lifetime maybe be ten thousand (or less) hours...barely enough to last a full year of 24x7 operation. Wait, how many of these are in a typical consumer computer? Even if it's just 20 or so (I truly have no idea) I don't get the numbers to add up. I've run, say, 6 computer-years worth of capacitors with no problems -- under the assumption above this would be 120 capacitor-years. Let's say at this point there's a 50--50 chance I would have experienced problems (this is not a rigorous survival analysis, of course, just something to get started). If a single capacitor has a lifetime of just one year, then "lifetime" must be defined as the solution to p^120 = 0.5 which is p=99.5 %. Is it correct that when one speaks of the "lifetime" of capacitors one refers to the point where 0.5 % of them no longer work? If "lifetime" means 50 % no longer work then I would seem to have been extraordinarily lucky (I'm the 10^-35 %!!) -- I don't hold that for very likely. What else am I missing?

TedDoesntTalk 4y ago

> a gaming PC may be only on a couple of hours a day a few times a week Never met such a gamer

zabzonk 4y ago
a gamer with a career and a life outside of "gaming"?
- Dylan16807 4y ago
  Doesn't mean you turn anything off.
coryrc 4y ago
More seriously though, lifetime of a component is generally halved for each ten Kelvin increase. So idling or lightly taxing the card at, say, 60°C instead of 90°C is a 8x longer lifetime alone.

anyfoo 4y ago

I corrected my text above to read "unreasonably short lifetime" now. In my experience, plenty of products from the 90s and older still seem to work absolutely fine despite their use of electrolytic caps. Yes, some of their parameters may have degraded, but it still seems to be within tolerance for their particular use in the design. It also depends on how exactly the caps are used. Caps close to a massive heat source, like a power resistor, are likely to dry out relatively quickly, while others, like those used as part of small signal filters, may last literal decades, with not much change in e.g. ESR or capacitance (and hence still conform to factory calibration pretty okay). Admittedly I don't have any hard data for that, only anecdotical experience with measuring ESR and capacitance and saying "this looks fine". The crucial point is that the "capacitor plague" affected much more than products which were on 24/7 or which had tight tolerances, to the point where it was very well visible to the "casual" consumer. Yes, we may not expect a server mainboard that was on for the last 20 years to still have its caps within spec, but wrt the capacitor plague, the gaming PC you mentioned would have been quickly affected, in an obvious manner, as well.

coryrc 4y ago
> while others, like those used as part of small signal filter I wouldn't expect those to be electrolytic, though. Electrolytic capacitors are mostly used when the exact characteristics are not important, like on a supply bus, where your qualifications are typically either "at least" or "no more" than some value. Tripling capacitance usually has no negative effect on a supply bus but huge change of doing signal conditioning, so something like that is more often going to be a capacitor that is intrinsically more stable.
- anyfoo 4y ago
  Good point, but in audio applications for example, for cross-over filters, bipolar electrolytic caps are common. The “calibration”, if any, may be very simple, though.

kqr 4y ago

> lifetime maybe be ten thousand (or less) hours...barely enough to last a full year of 24x7 operation. Wait, how many of these are in a typical consumer computer? Even if it's just 20 or so (I truly have no idea) I don't get the numbers to add up. I've run, say, 6 computer-years worth of capacitors with no problems -- under the assumption above this would be 120 capacitor-years. Let's say at this point there's a 50--50 chance I would have experienced problems (this is not a rigorous survival analysis, of course, just something to get started). If a single capacitor has a lifetime of just one year, then "lifetime" must be defined as the solution to p^120 = 0.5 which is p=99.5 %. Is it correct that when one speaks of the "lifetime" of capacitors one refers to the point where 0.5 % of them no longer work? If "lifetime" means 50 % no longer work then I would seem to have been extraordinarily lucky (I'm the 10^-35 %!!) -- I don't hold that for very likely. What else am I missing?