How long for capacitors??

How long should it take before one can be confident that capacitors that
have not yet gone bad aren't likely to do so?

I'm still mucking around with my KT7A Raid. It won't cold boot, it'll warm
boot after the CMOS is cleared. Making changes to the CMOS seems to be
something that pushes it through a cold boot, so that is a no go.

It's not a PS problem, unless the brand new 480 watt one I tried had the
same flaws as the V-Tech 300 Watt one that is in there. Brand new CMOS
battery didn't help.

Apparently it's now defaulting to being a 600 mHz machine, and unless fixed
will limp along that way forever. It's pretty stable though.

So I started looking at an old KT7 of roughly the same vintage that I ruined
years ago while attempting to remove the heat sink with the aid of a
screwdriver. I looked at the caps on it, and noticed that two of the larger
ones on that KT7, the 2200 uF ones at 6.3V, have domes on top. I don't see
any domes at all on the comparable capacitors of the KT7AR that's acting so
crippled. Then I look carefully at the smaller electrolytics on the KT7AR
and notice that some of them ...while not having domes, aren't exactly flat
anymore on top. It seems that a few of them have some unevenness along
their tops. Most of the comparable ones on the old KT7 are still as flat as
can be.

I also looked carefully at a DFI Mobo that ruined itself one fine day when
the fan died, nobody noticed for a while, and ...again (what an
embarassment), the screwdriver trick make a few of the runs on the
motherboard non-continuous. Crappy mobo anyway. Was very interesting,
though. ALL the 2.2K uF capacitors have domes on them....every last one of
them, and some of them have evolved some crud. The smaller ones look
perfect.

Well, I'm buying a new (ugh...store bought) computer, I don't have great
faith in the KT7AR anymore, although if it could be fixed up and made
reliable (it's not a real joy to have to clear the CMOS before every cold
boot), I'd be pleased.

I've started to practice unsoldering capacitors from the DFI board, with
iron, solder sucker, and fingers. When I get that down, I'll start doing
the same thing on the KT7; throwing away all the caps that have domes or
uneven tops. Then with the new computer in hand and running, I'll have a go
at replacing caps on the KT7A Raid, in the hope that a cold boot capability
along with something faster than 600 mHz can be achieved.

Are those caps on my KT7AR the real reason it doesn't want to do what it
should be doing? I have no idea, and only if replacing the ones that look a
little funny makes the problem go away will I be fair in suspecting caps as
the problem. Is it worth sending the thing to Homie and paying $70 just to
salvage this Mobo and get to keep on using PC133 RAM? Of course not, even
if it works afterward.

Exray wrote:
> How long should it take before one can be confident that capacitors that
> have not yet gone bad aren't likely to do so?
>
> I'm still mucking around with my KT7A Raid. It won't cold boot, it'll warm
> boot after the CMOS is cleared. Making changes to the CMOS seems to be
> something that pushes it through a cold boot, so that is a no go.
>
> It's not a PS problem, unless the brand new 480 watt one I tried had the
> same flaws as the V-Tech 300 Watt one that is in there. Brand new CMOS
> battery didn't help.
>
> Apparently it's now defaulting to being a 600 mHz machine, and unless fixed
> will limp along that way forever. It's pretty stable though.
>
> So I started looking at an old KT7 of roughly the same vintage that I ruined
> years ago while attempting to remove the heat sink with the aid of a
> screwdriver. I looked at the caps on it, and noticed that two of the larger
> ones on that KT7, the 2200 uF ones at 6.3V, have domes on top. I don't see
> any domes at all on the comparable capacitors of the KT7AR that's acting so
> crippled. Then I look carefully at the smaller electrolytics on the KT7AR
> and notice that some of them ...while not having domes, aren't exactly flat
> anymore on top. It seems that a few of them have some unevenness along
> their tops. Most of the comparable ones on the old KT7 are still as flat as
> can be.
>
> I also looked carefully at a DFI Mobo that ruined itself one fine day when
> the fan died, nobody noticed for a while, and ...again (what an
> embarassment), the screwdriver trick make a few of the runs on the
> motherboard non-continuous. Crappy mobo anyway. Was very interesting,
> though. ALL the 2.2K uF capacitors have domes on them....every last one of
> them, and some of them have evolved some crud. The smaller ones look
> perfect.
>
> Well, I'm buying a new (ugh...store bought) computer, I don't have great
> faith in the KT7AR anymore, although if it could be fixed up and made
> reliable (it's not a real joy to have to clear the CMOS before every cold
> boot), I'd be pleased.
>
> I've started to practice unsoldering capacitors from the DFI board, with
> iron, solder sucker, and fingers. When I get that down, I'll start doing
> the same thing on the KT7; throwing away all the caps that have domes or
> uneven tops. Then with the new computer in hand and running, I'll have a go
> at replacing caps on the KT7A Raid, in the hope that a cold boot capability
> along with something faster than 600 mHz can be achieved.
>
> Are those caps on my KT7AR the real reason it doesn't want to do what it
> should be doing? I have no idea, and only if replacing the ones that look a
> little funny makes the problem go away will I be fair in suspecting caps as
> the problem. Is it worth sending the thing to Homie and paying $70 just to
> salvage this Mobo and get to keep on using PC133 RAM? Of course not, even
> if it works afterward.
>

Have you ever tried to clear the CMOS, while the computer was still plugged
in ? On some motherboards, you can burn a small ORing diode. (And usually
the user manual will mention unplugging or switching off the computer, as the
first step in the clear CMOS procedure.) The diode has three legs, because in fact
it consists of two diodes in a single package. The schematic symbol for one,
looks like this (view in a fixed font like Courier):

|\ |
(Power 1 _____| \|__________
Source) | /| |
|/ | |
|___________ 3 (Load side)
|
|\ | |
(Power 2 _____| \|__________|
Source) | /|
|/ |


Your power supply, and its +5VSB related source might be connected to (1).
The CMOS battery, through a 1K resistor, is connected to (2).
The Southbridge and its CMOS logic block, connect to (3).

The diode prevents current from flowing in reverse, and into the CMOS
battery. The two diodes also "choose" the highest voltage power source.
The power supply on (1) has the highest voltage while the computer is
plugged in and +5VSB is being delivered. When the computer is turned off,
then (1) is at zero volts, so the battery on (2) wins. It gets to power the
Southbridge while the computer is not powered.

Now, the fun starts, if you clear the CMOS, while the computer is still
delivering +5VSB. For reasons unknown, some motherboard designs short (3)
to ground, to clear the CMOS (not all do this, but to be safe, you should
always unplug any computer you clear the CMOS on). If the power supply is
connected, that causes amps of current to flow from (1) to (3), burning the diode.

The symptoms would be:

1) You have a brand new CMOS battery, and it doesn't help.
2) Every time you switch off at the back of the computer, or unplug the
computer, the next time you use it, all the BIOS settings are reset.
And most likely, the BIOS clock will be reset to 1970 (or whatever
its reset-year happens to be).

Only one guy I posted this info for, ever managed to fix this. In his
case, he removed the old "K45" device and put a couple switching diodes
(like 1N914 or 1N4148 types). Ideally, you'd want a Schottky diode type
as a substitute.

How do I know it says "K45" ? I have three different model motherboards
here, and the same basic device is used on all of them. You need a magnifying
glass, to see the "K45" on the device. While a manufacturer
can use something else, they still need the functionality, so there should
be something similar on your motherboard. The device can be just about
anywhere on the motherboard, but you would hope the designer put it within
a few inches of the CMOS battery.

This is an example of a BAS40W-05. You want the -05 version, to get
the right "flavor" of diode connections. This document says you'd see
"K45" printed on the top of the three legged device, that is, unless
it is burned and illegible. Notice that the "marking" on the device,
does not match its part number - a practice that boggles my mind.

http://web.archive.org/web/200306292...cn_b_3_452.pdf

One replacement device costs $0.90. Or you can fashion an equivalent
solution, by soldering two ordinary diodes together, to make a
three-legged replacement.

http://www.digikey.com/scripts/DkSea...375549&Site=US

That would be my best guess, unless you've got a bad connection in the
battery socket.

Your "bad cap" problem is a totally separate issue. I don't think there is
any way to predict the failure date. The chemistry is temperature sensitive -
the hotter the environment, the sooner the failure. "Bad caps" fail in half
the time of regular caps. But there are other factors at play too - such as
whether the motherboard maker was cheap, and used the bare minimum number of
caps, with no design margin. Basically, the caps share "ripple current", which
is flowing to ground. When one cap fails, its share of the ripple current
will flow in the other caps, making their load heavier. That is why you may
see correlated behavior in the failures, and when one in a "clump" of caps
goes, the others may soon follow.

When you have empirical evidence of potential failures (like USENET postings
from fellow users), you can replace the caps ahead of time. Then, there is no
wondering later, about when they are about to fail.

At one time, "Homey" used to recap motherboards for $50, and if he was
still in the business, the $50 would be well spent. The replacement process
is complicated by how small they made the holes for the caps, and some
designers use an "interference fit", meaning the hole is barely larger than
the size of the capacitor lead. Those are a bugger to get out, and if that
is what they look like, getting someone else to replace all of them for
$50 would be a good deal (relative to the hours of cursing and swearing
you'd do, if you did it yourself).

Paul

Paul, thanks so much for your response. That little diode or gate really is
tiny. I found it on my scrap KT7 right near the mystery jumper, JP1, that
the manual doesn't bother to address, perhaps 4 inches from the battery, and
interestingly, very near a small array of 4 conventional diodes.

It's not impossible that I have at some time in the past cleared the CMOS
while power was still available to the mobo.

The CMOS does seem to maintain its memory even with power off. I know that
exactly because it won't cold boot, and of course those failures are
associated with new CMOS values that push toward a higher CPU clock rate.
That's not definitive, of course, but it's good circumstantial evidence.

It'll be a while before I do a serious root canal on this machine; I've
already had such stunning success this week off from work, I look forward to
abandoning this KT7AR for a while and using the files and settings transfer
wizard to get a brand new store bought machine up and running.

You would have had to mention 1N4148s as possible reasonable alternatives
for that K45. Damn...I have a drawerful of them, so if it turns out to be
the issue, I'll be able to wire around it.

The history of this problem is that another computer with a 2600+ froze up,
and from then on didn't post. I figured I'd check to see if the processor
was OK and put it into the KT7AR...and I got the same result. I did see,
once, a default on the new processor to 2000 mHz, but basically that was the
start of the problem in the KT7AR. I am starting to imagine that the 2600+
damaged the mobo, took out some 3.3 V lines or some grounds, and now that's
compromised the amount of power that can be gotten to any processor in that
socket. I can't yet figure out just how I'd find out if that is the
situation.

Of course one thing I could try would be to set the CMOS to 6x100 and do a
save and see if THAT provides a cold boot (I suspect it would). Oh well,
this computer stuff can't work anyway, we all know that.

Exray wrote:
> Paul, thanks so much for your response. That little diode or gate really is
> tiny. I found it on my scrap KT7 right near the mystery jumper, JP1, that
> the manual doesn't bother to address, perhaps 4 inches from the battery, and
> interestingly, very near a small array of 4 conventional diodes.
>
> It's not impossible that I have at some time in the past cleared the CMOS
> while power was still available to the mobo.
>
> The CMOS does seem to maintain its memory even with power off. I know that
> exactly because it won't cold boot, and of course those failures are
> associated with new CMOS values that push toward a higher CPU clock rate.
> That's not definitive, of course, but it's good circumstantial evidence.
>
> It'll be a while before I do a serious root canal on this machine; I've
> already had such stunning success this week off from work, I look forward to
> abandoning this KT7AR for a while and using the files and settings transfer
> wizard to get a brand new store bought machine up and running.
>
> You would have had to mention 1N4148s as possible reasonable alternatives
> for that K45. Damn...I have a drawerful of them, so if it turns out to be
> the issue, I'll be able to wire around it.
>
> The history of this problem is that another computer with a 2600+ froze up,
> and from then on didn't post. I figured I'd check to see if the processor
> was OK and put it into the KT7AR...and I got the same result. I did see,
> once, a default on the new processor to 2000 mHz, but basically that was the
> start of the problem in the KT7AR. I am starting to imagine that the 2600+
> damaged the mobo, took out some 3.3 V lines or some grounds, and now that's
> compromised the amount of power that can be gotten to any processor in that
> socket. I can't yet figure out just how I'd find out if that is the
> situation.
>
> Of course one thing I could try would be to set the CMOS to 6x100 and do a
> save and see if THAT provides a cold boot (I suspect it would). Oh well,
> this computer stuff can't work anyway, we all know that.
>

When I read your symptoms, maybe I got it wrong. I thought you weren't able
to keep the settings. If it is keeping its settings, then the diodes should be
OK.

If you only have stability with a low core frequecy setting, yes, that can be
a Vcore power issue. You could run a copy of CPUBurn or the like (not Prime95,
because it would probably error out and stop instantly). That would load the
processor 100% and make it draw max power. If there was a problem with the
Vcore regulator, the voltage would drop more than 200mV from the load.
While Vcore is *designed* to droop a little bit under load, there is
an acceptable mount of droop (there are load lines in processor datasheets
that show the limits allowed). Failing caps can compromise the ability of
the regulated Vcore circuit, to do its job.

While a store bought computer has a comforting ring to it, remember that
inside, it is manufactured by the same kind of people who made your
current board. Yes, you get a warranty, you get that wonderful tech support
on the phone, but underneath, you still have the same potential for issues.
One reason I like DIY systems, is they are usually easier to fix up, than
the store bought system (I'm referring to that restore CD they give you).

I was defeated by an early build I did. I bought a motherboard that had
the world's worst Northbridge on it. My graphics card could never be made
stable with that motherboard. It sits in my "hall of shame" (broom closet).
I spent three weeks testing and tweaking, but could not fix it. Since it
"kinda worked", and I was a noob, I kept it, thinking I didn't have any
alternatives. Now, I realize I should immediately return stuff like
that - live and learn.

Paul

"Exray" <> wrote in message
news:RM-dnZ4pjdiuIgvYnZ2dnUVZ_h-...
> How long should it take before one can be confident that capacitors that
> have not yet gone bad aren't likely to do so?

To be fair, most of the KT7s in use nowadays have most likely had caps
replaced at some point - it's very easy, just make sure you get a
temperature controlled iron (50W or so) as the groundplane on the board
tends to soak up a huge amount of heat and makes them awkward to remove.
Once you've got them out, don't bother to clear out the holes - simply trim
the legs on the new caps so they are about 15mm long, line them up with the
solder-filled holes and push gently down whilst heating the board from the
solder side - do this one leg at a time and the legs will just slip through,
it is *so* much quicker than clearing out the little buggers Last mobo I
did this way with 8 caps took about 20 mins total time!

Hellraiser..............>

So I downloaded and installed CPUBurn, and also Speedfan, which does some
convenient real time readout of things like Core Voltages.

At 100% of CPU, the core and +5 drooped a tiny bit, but not significantly.

So then I went into the bios, set it at 6 x 100, and it posted OK with no
CMOS clearing needed.

Then I got ambitious, set it at 10 x 100, and -- Splat; same problem.
There's probably someplace in between that would work; but it's like taking
two spark plug wires off and then seeing how well the carburetor can be
tuned to optimize the zero to 60 time. Knowing the plug wires are off and
being able to reconnect them would be the better solution, but for that
little problem of knowing.

If the machine boots and runs OK at, say, 750, or even 5 x 133, I'm not sure
I'd have learned a hell of a lot about what to fix.
"Paul" <> wrote in message news:en713k$iqn$...
> Exray wrote:
>> Paul, thanks so much for your response. That little diode or gate really
>> is tiny. I found it on my scrap KT7 right near the mystery jumper, JP1,
>> that the manual doesn't bother to address, perhaps 4 inches from the
>> battery, and interestingly, very near a small array of 4 conventional
>> diodes.
>>
>> It's not impossible that I have at some time in the past cleared the CMOS
>> while power was still available to the mobo.
>>
>> The CMOS does seem to maintain its memory even with power off. I know
>> that exactly because it won't cold boot, and of course those failures are
>> associated with new CMOS values that push toward a higher CPU clock rate.
>> That's not definitive, of course, but it's good circumstantial evidence.
>>
>> It'll be a while before I do a serious root canal on this machine; I've
>> already had such stunning success this week off from work, I look forward
>> to abandoning this KT7AR for a while and using the files and settings
>> transfer wizard to get a brand new store bought machine up and running.
>>
>> You would have had to mention 1N4148s as possible reasonable alternatives
>> for that K45. Damn...I have a drawerful of them, so if it turns out to
>> be the issue, I'll be able to wire around it.
>>
>> The history of this problem is that another computer with a 2600+ froze
>> up, and from then on didn't post. I figured I'd check to see if the
>> processor was OK and put it into the KT7AR...and I got the same result.
>> I did see, once, a default on the new processor to 2000 mHz, but
>> basically that was the start of the problem in the KT7AR. I am starting
>> to imagine that the 2600+ damaged the mobo, took out some 3.3 V lines or
>> some grounds, and now that's compromised the amount of power that can be
>> gotten to any processor in that socket. I can't yet figure out just how
>> I'd find out if that is the situation.
>>
>> Of course one thing I could try would be to set the CMOS to 6x100 and do
>> a save and see if THAT provides a cold boot (I suspect it would). Oh
>> well, this computer stuff can't work anyway, we all know that.
>
> When I read your symptoms, maybe I got it wrong. I thought you weren't
> able
> to keep the settings. If it is keeping its settings, then the diodes
> should be
> OK.
>
> If you only have stability with a low core frequecy setting, yes, that can
> be
> a Vcore power issue. You could run a copy of CPUBurn or the like (not
> Prime95,
> because it would probably error out and stop instantly). That would load
> the
> processor 100% and make it draw max power. If there was a problem with the
> Vcore regulator, the voltage would drop more than 200mV from the load.
> While Vcore is *designed* to droop a little bit under load, there is
> an acceptable mount of droop (there are load lines in processor datasheets
> that show the limits allowed). Failing caps can compromise the ability of
> the regulated Vcore circuit, to do its job.
>
> While a store bought computer has a comforting ring to it, remember that
> inside, it is manufactured by the same kind of people who made your
> current board. Yes, you get a warranty, you get that wonderful tech
> support
> on the phone, but underneath, you still have the same potential for
> issues.
> One reason I like DIY systems, is they are usually easier to fix up, than
> the store bought system (I'm referring to that restore CD they give you).
>
> I was defeated by an early build I did. I bought a motherboard that had
> the world's worst Northbridge on it. My graphics card could never be made
> stable with that motherboard. It sits in my "hall of shame" (broom
> closet).
> I spent three weeks testing and tweaking, but could not fix it. Since it
> "kinda worked", and I was a noob, I kept it, thinking I didn't have any
> alternatives. Now, I realize I should immediately return stuff like
> that - live and learn.
>
> Paul

"Exray" <> wrote in message
news:4vKdnVx1r-...
> So I downloaded and installed CPUBurn, and also Speedfan, which does some
> convenient real time readout of things like Core Voltages.
>
> At 100% of CPU, the core and +5 drooped a tiny bit, but not significantly.
>
> So then I went into the bios, set it at 6 x 100, and it posted OK with no
> CMOS clearing needed.
>
> Then I got ambitious, set it at 10 x 100, and -- Splat; same problem.
> There's probably someplace in between that would work; but it's like
> taking two spark plug wires off and then seeing how well the carburetor
> can be tuned to optimize the zero to 60 time. Knowing the plug wires are
> off and being able to reconnect them would be the better solution, but for
> that little problem of knowing.
>
> If the machine boots and runs OK at, say, 750, or even 5 x 133, I'm not
> sure I'd have learned a hell of a lot about what to fix.


I'd still say capacitors - the tiny voltage droop is normal as the cpu is
under load, but the caps themselves filter out all the crap on the supply
lines, if they start to fail then you get all sorts of glitches and hiccups
getting through and it is those that cause either instability, or failure to
run at higher clock frequences. Either scrap the board, or replace the caps
and try from there otherwise you'll be wasting a whole lot of time

Hellraiser..............>

Thanks Hellraiser. I'm headed toward scrapping the board, the whole machine
is all backed up. The future activities with this machine will definitely
be a "spare time" activity. What I'd HATE to do is replace the caps and
discover that one of the other of the jillion parts on the board is at
fault.

"Exray" <> wrote in message
news:. ..
> Thanks Hellraiser. I'm headed toward scrapping the board, the whole
> machine is all backed up. The future activities with this machine will
> definitely be a "spare time" activity. What I'd HATE to do is replace the
> caps and discover that one of the other of the jillion parts on the board
> is at fault.

True, but it's only an hours worth of work, and if you got some with
non-flat tops already, coupled with the fact that Abits are notorious for
bad-cap problems, it's a no-brainer I've done loads of motherboards from
Gigabyte, Abit, MSI and others and it is almost always caps that are the
problem.

Hellraiser................>