I've been seeing a lot of motherboard capacitor failures for over a year now, affecting Socket 370 and 478 generation Intel motherboards, as well as AMD contemporaries. Affected systems may present with:
The problems are typically progressive, over days to months. All the motherboard brands I commonly see (DFI, Gigabyte, Jetway and MSI) are affected; other brands reputed to have this problem, e.g. Abit, I don't see here. Often it's the capacitors around the processor, obscured by the power supply, that are affected - so if checking for this, remove the power supply if need be.
What to look for
Electrolytic capacitors store charge locally on the motherboard, for a faster supply of current that would be possible from the distant power supply (on nanosecond timescales, that distance can matter). They are small cylinders that contain foil and insulator layers rolled up in liquid goo inside. When stressed, the goo may break down to release hydrogen, which bulges the tops of the capacitor until it leaks, or alternatively they can leak from the bottom. Sick ones may look like this:
For every fifty or so affected motherboards, I've seen similar pattern of failure in power supplies...
...and graphic cards:
It's possible to replace these defective capacitors, even when these are in multi-layer boards where soldering the middle layers would seem a challenge. Keep your repair slip, in case other capacitors that looked OK should fail later. As repair shops do more of these, they are more likely to replace all caps of an affected size and value, rather than just the obviously distressed ones. So far I've had two motherboards that did not survive repair; one was replaced via warranty swap (2 year warranty, luckily) while the other precipitated an "upgrade opportunity".
The cause
I've read articles that speculate on bad electrolyte solution, based on a stolen formula that was incomplete. And I thought; these capacitors have been around in old radios etc. for decades; why would any factory have to hunt for a formula to make the goo? Surely it's a well-known non-secret by now?
A bit more reading suggested what the problem might be. We've subjected these capacitors to high frequencies and high currents, but not at the same time - until modern processors boosted both current draw and ripple frequency to new highs. And so it is that a new "low ESR" grade of capacitor is needed for this purpose, and the electrolytic solution used in those capacitors is likely to be more exotic and proprietary.
Perhaps Moore's Law will falter on nuts and bolts issues like this, if not heat dissipation problems. Normally, each processor fabrication size shrink has resulted in lower current and heat that allows higher clock speeds and more transistors to be added within the chip. But the last (Prescott) re-fabrication of Intel's processors has left us with chips that run hot; so we have noisy fans, overheating cases, and bulky new case designs to shunt the heat out of the box.
(C) Chris Quirke, all rights reserved; March 2005