PC motherboard failure information

ASRock 939Dual-SATA2

After about eight years use of these boards I have seen various problems start to appear. Previously I had replaced all the large 3300uF 6.3V capacitors in the CPU VRM circuitry which cured a failure to POST problem. Now one particular PC using this board would POST but would display a blank screen or would become stuck on the Windows logo during Windows XP startup or, if it did start, would later show instability. Tests on memory, CPU etc passed so I reinstalled Windows XP but on completion the problem remained.

I decided that it was a capacitor problem and the OST RLS 1000uF 6.3V capacitors needed replacing, not a decision to be taken lightly as there are 25 of them. An interesting discovery was that the M1695 Northbridge heatsink on the problem board was running hot, so hot that I could not keep my finger on it for very long. The M1695 Northbridge has a cluster of capacitors below it supplying its power and two of these were also getting very hot. After some further checking I found that there are three pairs of capacitors providing power to the M1695 NB and the M1567 SB. Each of these consists of a OST RLS 1000uF 6.3V low ESR capacitor and a TEAPO 100uF 16V standard ESR capacitor connected in parallel which was a surprise. Using an ESR meter the 100uF capacitors were measuring over 1ohm compared to the OST capacitors which, with the odd exception, measured around 0.04 ohms. After replacing just the 100uF capacitors with new equivalents the board ran correctly and the NB heatsink was much cooler.

As far as I can tell these are the capacitors which tend to fail in this area, probably because of their proximity to the NB heatsink.

CE91 1000uF, CE7 100uF (M1695)
CE82 1000uF, CE83 100uF (M1695 ?)
CE34 1000uF, CE30 100uF (M1567 ?)

CE40,CE41,CE42,CE43 & CE44 are all OST 1000uF 6.3V presumably for the DDR ram. Sometimes one of these will bulge and its ESR will increase but, because they are all connected in parallel, it does not seem to have much effect on board reliability. I have found that the other OST 1000uF capacitors on the board do not tend to fail.

A second 939Dual-SATA2 motherboard would start up correctly but would not shutdown from Windows and would remain powered on with no display. This was found to have a very high ESR value for the capacitor CE30 (> 3 ohms) and CE7 & CE83 were higher than they should be. The associated 1000uF capacitors were all found to be ok.

Another board had the original failure to boot Windows problem and the NB heatsink was very hot but it seems that the damage had already been done to the M1695. It would only boot if I pressed down on the NB heatsink, I assume because the solder connections on the BGA package had become disconnected as a result of the severe overheating. The lesson to be learnt here is to check these capacitors if the NB heatsink is running hotter than normal.

The conclusion is, failure to boot up or shutdown reliably and/or instability may be caused by the capacitors mentioned and replacing them may fix the problem without the wasted time and expense of a full re-cap of the board.

AMD K7 processors

It was quite common to overclock AMD K7 processors and this would be fine if the processor was cooled sufficiently. If not then the result would be a dead CPU and possible damage to the motherboard too. There is a good chance that a used K7 processor or motherboard obtained now would have been overclocked in the past. Recently I was testing K7 processors in a motherboard believed to be working and when the ATX power switch was activated the fans ran for a very short time before the motherboard powered off. Luckily, the motherboard had built-in protection from a processor overheating.

When a K7 processor is damaged internally by overheating the result is a lowering of the resistance between VCore and GND. If this resistance could be measured then potential motherboard damage by using a faulty processor could be prevented. It would be tricky to use a multimeter to measure resistance between a VCore pin and a GND pin on the processor itself but as VCore is connected to output of the VRM it could be calculated by making two resistance measurements from the VRM output to GND. One measurement is made with the processor fitted to the motherboard and the other with it removed. To do this the motherboard/PC must be turned off and only if the processor passes the test would the motherboard be powered on.

Either

(If the underside of the motherboard is accessible) Measure across pins of the large capacitors near to the processor socket which smooth the VCore output.

or

Measure between one soldered pin of a choke (ring with copper wire wrapped through it) near to the processor and a motherboard mounting screw.

Always touch the same probe to the same place on the motherboard and wait a few seconds for the reading to settle. If a reading close to zero is found for RM2 it indicates damage to the motherboard VRM, otherwise, if a reading close to zero is then found for RM1 it indicates that the processor is shorted and useless.

Using the formula for resistances in parallel

1 / RM1 = 1 / RM2 + 1 / RP

where

RM1 = measured resistance with processor fitted
RM2 = measured resistance with processor removed
RP = internal resistance of processor

by rearranging

RP = (RM1 x RM2) / (RM2 - RM1)

Sensible values of RP for different K7 processor cores

Spitfire 205 ohms
Thunderbird 95 ohms
Palomino 37 ohms
Thoroughbred 18 ohms

e.g. For three Thunderbird processors, of which only one is working, measured on a Asus A7V333 motherboard

a) RM1 = 1.3 ohms RM2 = 33.5 ohms, RP = 1.35 ohms
b) RM1 = 9.3 ohms RM2 = 33.5 ohms, RP = 12.87 ohms
c) RM1 = 25.4 ohms RM2 = 33.5 ohms, RP = 105 ohms

If the calculated RP value is more than +/- 20% from the expected value then the processor is likely to be damaged. However, this measurement cannot prove with certainty that a processor is working.

home contact