Why Cap.'s fail and cures.                  

TOPIC Electrolytic Capacitor life. (short lived)  Modern Cap.'s today are water free.( organic solvents.)
HEAT is the killer of all electronics, and CAPS are the worse of them all (not counting dag goomed batteries)
A PC is spec'd as 95F (35C) max room temp so the cheap and  lame PCs with no case fans get hot, 120 to 150f (65c)and this will stress those caps (less life) so life is 30/10=3 or times less. (read A LOT)
Best coverage of all is in the  wiki on topic. (see "Service life") and low ESR caps are worse due to  more water inside, (with  the borax)
All of these caps dry out, there is no 100 year life or more like seen on quality transistors and diodes, or resistors. Even many top EEPROM ave cell charge life of 200 years. Just the programming lasts that long (alone)
This parts have an Achilles heal. (the electrolyte part) these parts are what I call WEAK TECHNOLGY and that means far better testing must be done. (for sure incoming inspections)
The other factor is internal heating from huge ripple currents across the cap pins, like seen on GPU or PSU devices.
Notwithstanding the for sure made in 2005  .Cloned caps failing, stolen forumas, et all.  (or just pure  crap companies new selling 1 cent parts)
Cap's seen to leak, crack or bulge are all 100% bad,  (a no brainer any kid can see)
The seals fail. ( the star cross stamps) are there to crack and throw the acid why high saving the PCB  from damage. (the acid will eat up any PCB  fast) boric acid or the like.
Some of these caps are from makers  that are only 2 years old, and no experience like NEC, PANASONIC and Nichicon top brands,  do not use unknown brands , gob smack me.
Bad seals bad electrolyte are common on this junk products.
(even stolen formula's in fact one formula stolen was incomplete and dead wrong,,,and all are  trade secret, for sure Panasonics great products)
The makes of sad caps do not test them at all for service life nor ISO9000 quality audit them at all. (per spec, or using the automotive spec on caps and test methologies)
Not only this but the makers of caps do not even do the 105C, 1000  hour, shelf life test, to make sure that is ok.  (nor the FIFO tests below) cheap is as cheap does.

Some CAPS go to 200,000 hours  hot.  some only 1000 hours.

At Mouser.com (great filters for sure) there are 5000 types of ALum solid organic polymer  sold there. 20,000 hours full 105c temp heat is their best. 2.2years life, but at lower temps vasty more 5 times more. (search for 65c is max internal air temp)
shows 63k hours at 75c life, 7.2 years, this is not end of life actual it is when specs are way off,  and most caps are installed 2 time more than needed for say brute force bulk ripple caps. (hard nut job)
What matters most is not leaking, not exploding nor drying out and going to 0 uF. Dead. 
The life is presumed to be 15 years here, that means far more. (in the real world no running at 105C)

Pros of polymer electrolytic capacitors:  (they cost more but quantity discounts of over 66% is normal)   20cents each is possible  but not  2 cents by China no-name maker. ( 25 caps total is $5 less saved per motherboard)

  • Lower ESR/higher allowable ripple current
  • No or way less dry-out behavior (unlike aluminum wet capacitors) nor leak and leaks can in fact wreck any $300 motherboard , a serious issue, and full replacement.
  • Higher expected lifetime/load life
Last and best is this rule,  lifetime doubles for every 10 degrees C below rated temperature. Arrhenius equation

 (yes heat kills)  (40c delta /10 = 4) so doubles 4 times.320k hours or 36 years,  as you can see they are far better. (life)

I once worked for a company (line ET tech) that or BUYER decided that USA transistors and caps and diodes were too expensive ! and a nutter him, got them from Hong Kong (clones) every one failed on the factory floor or field. (bankruptcy fast) circa 1976.
Just one man wrecked a whole company, the Chief engineer was not allowed to do his job.
Avoid: these caps. (RMA fails max) Use only tier 1 parts,. or go broke fast.
  • Jamicon
  • CapXon
  • G-Luxon
  • Su'scon
  • Lelon
  • Ltec
  • Jun Fu
  • Fuhjyyu
  • Evercon

Testing 1,2,3:

Testing the maker does not test there product ,so  this is why they ship crap, for 2cents each. (testing costs money for sure if having to scrap a full lot of parts, and pretend to fix  the problem next)
There are 2 tiers of testing the junk cap maker and the poor OEM maker of PC or PSU or GPU cards, that must protect there product line and for sure reputation, seems Dell was asleep. (seems now most OEM did)

It is  true they (PC maker) did not take samples of every lot and burn them all in at 85C (cheap grade) and used to filter noise 60Hz 1/2wave ripple for 31 days, just sample with the matching lot,  held in quarantine, a LIFO , last in first out.
If the test even 1 sample fails  then  the whole bad lot is rejected and sent back to CHINA (in the contract are those ISO9001 like words in fact) If you are the cap maker the lot is crushed, and recycled raw materials, (aluminum)
One could even force the China cap, maker to do the burn-in, per above and not ship CRAP. (by negotiating on huge orders, or learn to disqualify them as  supplier, or have them tested by 3rd party lab. Lots of choices here...)
In many a company I worked in we had whole sections of engineering just for parts qalifications, thermal stressing , burn-in, vibration testing, and more, if parts failed?,  off the list they go.  5 cent parts be DAMNED !
There are endless crap parts "components" made now (unstoppable for sure!), it is the PC makers JOB (PSU,/GPU etc) to test the parts before theyu are ever USED in PRODUCTION. (or go broke fast you will !)
Just in time (JIT) be damned you need  a 60day supply of parts for a 30day long production, one 1/2 still in burn-in, just like we do on military grade parts. (burned and cycled "clocked" fast) HOT.
The real truth is buy a PSU and GPU from USA stores only and top brands and they all will be good today. (most are) And not that one for $15 on fleabay , ok,? for sure made in 2005, wow.
Now the clincher, if you test all those full lot parts correctly they are damaged goods. (but the FIFO method works,  those samples are never sold and the lots never stressed nor there life spans shorted from stress testing)
LIFO  is king, in my book.  (see this work me)  at 105C (the correct parts to use  are those for sure)

One more trick on LIFO,  is the pipeline, as it is called, we had  plant in Korea and Singapore. and the shipping pipe line was about 1 month to CONUS.
So the samples were flown to USA fast in 1 day. tested for 30days at 105C and as the ship arrived, the lots bad pulled and destroyed.  (cute no?)  Never forget using pipelines to get the job done.
We also did the LIFO in Korea later, and they did their  own tests/ scraping, and we did not ship crap to our buyers ever.  (or held the part lots for 30 days, there and saved shipping carp to USA) We did it right and never shipped crap.
We tested the hell out of parts some from -55c to 100C.  full stress and full clocking of parts, or in case of CAPS rippled tested max.  (samples only but MIL testing is ALL parts fully tested before stamped)

version 1.  1 -20-2020