Caps on motherboard:
2 I.Q. 1500uf @ 10v, both vented out the bottoms. Replaced with Nichicon 1500uf @ 10v PM series.
8 Choyo 1500uf @ 6.3v, all 8 (!) vented out the bottoms. Replaced with Nichicon 1800uf @ 6.3v PL series.
4 Teapo 1000uf @ 6.3v. Good physical shape. Decided to ignore. While Teapos frequently fail with no external signs (and I've replaced plenty of perfectly healthy looking ones before), these are tiny ones, which I haven't seen fail...
3 S.I. 1000uf @ 10v. Good physical shape. Decided to ignore. I haven't seen S.I. caps fail (although someone who has is free to correct me), and again these are small, which I also don't tend to see fail very often.
Yes, I left 1000uf caps on the board. On most boards I'd replace all of them as a preventive measure, but these are physically smaller than any ones I've seen failed before, are different brands than the failed ones, show no physical signs, at least in the case of the S.I. ones don't look like low-esr ones, and it's an old motherboard with no plans of a future as anything other than a node in my renderfarm. Plus these are so tiny that their replacements are probably fifteen (or more) times larger, and would be a pain to make fit.
Board history:
This board was used as a number-crunching box with a p3-733, for doing a study on reducing the problem of solving the historical transposition cipher using hamiltonian paths, specifically the tradeoffs between a couple parameters, the probability of solving it, and the time it takes to solve it Output was pretty rainbow-colored 3d surfaces. When the box became unstable, bad caps were noted, but the box couldn't be taken out of service, so the clock speed was lowered until it was stable again. Repeat about monthly until the bus speed is half of what it should be, then eventually doesn't turn on, at which point it finally is replaced.
Due to this drawn-out failure process, all ten of the larger caps on the board vented dramatically before the board finally stopped POSTing. It's not too clear from the photos, but probably a good portion of a teaspoon of electrolyte was scraped off the board with paper towels and alcohol. Many parts of the VRM are corroded, but nothing that looks enough to keep it from functioning, after soaking them in alcohol at least...
When the board was finally pulled, it would turn on, and that's it. Never initialized video or anything else.
Replacement: Straightforward. Only 3 of the leads (of 20) took a 2nd try with the desoldering gun, all the others went instantly. Only one needed a quick poke with a cold pick, due to having a tiny solder bump left. This is the way they should be! none of the horrible mess with hot picks and 140 watt guns that gainward board took.
Before:
Note the visible electrolyte, exposed rubber plugs, domed tops, and random angles.
One of the many electrolyte splatters across the board.
Underneeth all the caps. The two to the side have already been replaced with shiny nichicons. (d'oh! that filename was the same as the photo of all the new caps I got. clobbering happened. oh well)
After:
Mmmm, shiny! They're all bent slightly outwards, to clear the cpu.
Small caps being ignored. Not the very small ones in the foreground, but the almost-very small ones behind them. Note their size relative to their replacements... even the old 1500uf 10v I.Q. ones are many times larger than the 1000uf 10v ones!
Testing rig. Only have a cheap cpu (celery 667 in a slotket) and a bit of ram in for testing... the small pliers function excellently as a power switch.
and...
nothing.
Powers on, but still never initializes video or does anything else.
Conclusion: While a bit less graphic than the "why you should never ignore bad caps" thread, the board is toast. Continuting to use it until its eventual hard failure probably caused something else to die.
It's theoretically possible one of the 1000uf caps I didn't replace is to blame, but I doubt it. I cleaned the electrolyte-caused corrosion up pretty completely, so don't think that's it either. Chances are it's just dead.
Now one of these days I need to actually put together the renderfarm, eh?
--Randy
2 I.Q. 1500uf @ 10v, both vented out the bottoms. Replaced with Nichicon 1500uf @ 10v PM series.
8 Choyo 1500uf @ 6.3v, all 8 (!) vented out the bottoms. Replaced with Nichicon 1800uf @ 6.3v PL series.
4 Teapo 1000uf @ 6.3v. Good physical shape. Decided to ignore. While Teapos frequently fail with no external signs (and I've replaced plenty of perfectly healthy looking ones before), these are tiny ones, which I haven't seen fail...
3 S.I. 1000uf @ 10v. Good physical shape. Decided to ignore. I haven't seen S.I. caps fail (although someone who has is free to correct me), and again these are small, which I also don't tend to see fail very often.
Yes, I left 1000uf caps on the board. On most boards I'd replace all of them as a preventive measure, but these are physically smaller than any ones I've seen failed before, are different brands than the failed ones, show no physical signs, at least in the case of the S.I. ones don't look like low-esr ones, and it's an old motherboard with no plans of a future as anything other than a node in my renderfarm. Plus these are so tiny that their replacements are probably fifteen (or more) times larger, and would be a pain to make fit.
Board history:
This board was used as a number-crunching box with a p3-733, for doing a study on reducing the problem of solving the historical transposition cipher using hamiltonian paths, specifically the tradeoffs between a couple parameters, the probability of solving it, and the time it takes to solve it Output was pretty rainbow-colored 3d surfaces. When the box became unstable, bad caps were noted, but the box couldn't be taken out of service, so the clock speed was lowered until it was stable again. Repeat about monthly until the bus speed is half of what it should be, then eventually doesn't turn on, at which point it finally is replaced.
Due to this drawn-out failure process, all ten of the larger caps on the board vented dramatically before the board finally stopped POSTing. It's not too clear from the photos, but probably a good portion of a teaspoon of electrolyte was scraped off the board with paper towels and alcohol. Many parts of the VRM are corroded, but nothing that looks enough to keep it from functioning, after soaking them in alcohol at least...
When the board was finally pulled, it would turn on, and that's it. Never initialized video or anything else.
Replacement: Straightforward. Only 3 of the leads (of 20) took a 2nd try with the desoldering gun, all the others went instantly. Only one needed a quick poke with a cold pick, due to having a tiny solder bump left. This is the way they should be! none of the horrible mess with hot picks and 140 watt guns that gainward board took.
Before:
Note the visible electrolyte, exposed rubber plugs, domed tops, and random angles.
One of the many electrolyte splatters across the board.
Underneeth all the caps. The two to the side have already been replaced with shiny nichicons. (d'oh! that filename was the same as the photo of all the new caps I got. clobbering happened. oh well)
After:
Mmmm, shiny! They're all bent slightly outwards, to clear the cpu.
Small caps being ignored. Not the very small ones in the foreground, but the almost-very small ones behind them. Note their size relative to their replacements... even the old 1500uf 10v I.Q. ones are many times larger than the 1000uf 10v ones!
Testing rig. Only have a cheap cpu (celery 667 in a slotket) and a bit of ram in for testing... the small pliers function excellently as a power switch.
and...
nothing.
Powers on, but still never initializes video or does anything else.
Conclusion: While a bit less graphic than the "why you should never ignore bad caps" thread, the board is toast. Continuting to use it until its eventual hard failure probably caused something else to die.
It's theoretically possible one of the 1000uf caps I didn't replace is to blame, but I doubt it. I cleaned the electrolyte-caused corrosion up pretty completely, so don't think that's it either. Chances are it's just dead.
Now one of these days I need to actually put together the renderfarm, eh?
--Randy
Comment