Cooked Caps?

[calzap]

A Dell Dimension 4500 desktop, bought new in 2002,  recently began locking-up or just going dead for no obvious reason.  I opened it, and the reason for the failure was clear.   The photos below show three failed caps on the motherboard.  I don't think this is part of the infamous, 12-million-computer, bad-cap fiasco that Dell covered-up for a while.  Those computers were made May 2003 through July 2005.  However, the caps were all Nichicon 3300 uF, 6.3 V, 105C.   Nichicon caps were the main brand in the bad-cap fiasco, so maybe this was a harbinger.  In addition, the bad-cap computers usually failed within three years, often after a few hours of service.  This computer operated continuously for 12 years.

Mostly, I think it's an example of poor thermal engineering exacerbated by the computer spending the last couple of years running relay boards for valve control in a building where temperatures could hit 41C in summer.  In the photos below, the three failed caps are all within 10 mm of the CPU.  They are between the fan and the CPU, but the lowest reach of the fan blades is 35 mm above the top of the caps, so they were in a little valley between the CPU and the case, and didn't benefit much from the fan.  It wasn't the best design for cooling the CPU either although there was a plastic shroud to deflect air down toward the  CPU on the opposite side from the caps.  This shroud might have caused a slight back-draft across the top of the CPU toward the caps, but it would take an airflow study to confirm it.

No other caps appear to have failed, and there were a lot of them elsewhere on the motherboard.  And the three failed caps comprised all of the caps next to the CPU.  I really can't complain too much because the computer did last 12 years of continuous operation, which is better than most, but it would have survived longer with a better thermal design.  I'll probably salvage some of the components and toss the rest.  An equivalent new computer is only US$330.  The caps could be replaced for less than this, but something else would probably fail in short order.  Not exactly BER, but close.

Have also had a hard drive failure in another computer doing the same function in the same building.   Anyone have experience with computers in warm environments and can make a recommendation on those that survive best?  Air conditioning the room where the computers live is not an option; it is well ventilated with outside air, but that can be 40C on hot days.  Have never had a CPU or memory problem though.   There is no catastrophe in this application if a computer failure is caught within a 1-2 days, which it always has been.  The computers are controlling irrigation valves, and the plants can survive a while without being watered.  However, it is a PITA because it seems to happen on weekends or when I'm busy with other projects. There is usually a standby computer available.  Worst case, but labor intensive, the  valves can be operated manually until a computer is fixed or replaced.

Mike in California

[mariush]

Between 2002 and 2004 (if i remember correctly), Nichicon had a manufacturing problem with a few series they made (HM and one or several others).. in simple terms they overfilled the capacitors with electrolyte.
The capacitors in those series with man. dates between those years (2002 until around june 2004 i think) have a high chance of dying due to heat more than other brands.

It's not related to capacitor plague, separate issue. At least they admitted the issue unlike UCC/NCC with their KZG/KZJ series (and a few others) capacitors that fail often without even bloating or leaking electrolyte... they just discontinued the series.

[jmoreland79]

Realistically you have a mid to  low-end consumer grade PC designed to run in a 25C-30C environment, so I don't think it's fair to call it a bad thermal design.  Inside any enclosure, you're going to have regions that will run hotter than others, and there's no way around that without an elaborate cooling system.  In fact, if the PC was really running for 12 years continuously, you got 100,000+ hours out of caps rated at 2,000 hours at 105C, so it did pretty well.

As for the hard drive failure, that has moving parts and will fail without fail (pun intended).  There really isn't much you can do to help those along.

If you're concerned about PCs in a warm environment, stay away from compact cases.  Stick with something like a mid-tower that will allow for more generous airflow through the chassis.  That said, it doesn't seem like you're in an extreme environment, so it doesn't make much sense to spend a lot extra on cooling.  Especially since you aren't running anything mission critical.  You can have the best cooling solution and end up with some oddball failure.  Consumer level PCs just aren't built to last that long, though they often do.

[tautech]

I have replaced the caps in a very similar situation and returned the MB to service with no problems.
In my case, caps were so bad the PSU would only fire up for a second or so then shutdown as the short was so bad.
Caps could burn your fingers in seconds and what surprised me was they were still intact and un-exploded.
They were however noticeably bulged.

[calzap]

I considered replacing the caps, but decided in a 12-year-old computer, something else would go out real soon.  So I removed the drives, PSU and memory and sent it to computer heaven with fond remembrances.  The new replacement, a Lenovo H series, cost US$260 and arrives in two days.  The Lenovo will be put to work for a few days making sure it can control the relay boards and survive the summer heat, then will go on the shelf as a standby.

Mike in California

[NiHaoMike]

If the rear fan has a thermal sensor, solder a piece of wire across the sensor and the fan will go full speed.