Capacitors on graphics cards...

On Wed, 27 Aug 2008 16:02:12 +1200, "~misfit~"
wrote:

Somewhere on teh intarweb "kony" typed:
On Wed, 27 Aug 2008 00:34:41 +1200, "~misfit~"
wrote:

Ok, so I have a couple Leadtek WinFast A280 LE AGP 8x VIVO 128MB
GeForce ti4200's. I bought them for my machine and my girlfriends
(back when I had one) and they cost over NZ$300 each new. They were
a significant investment at the time, costing twice as much as the
T'bredB Athlon XP CPUs they ran with.

My ex's card failed before we split and I put it on the shelf and
bought her a 7600GS to replace it. Mine failed not long afterwards
and also went on the shelf. I couldn't bring myself to throw them
away considering what they cost. Actually, only one is VIVO, the
other didn't need to be VIVO and was ~$40 cheaper.

The other day I was having one of my periodic clean-outs and I had a
good look at them. One of them had a capacitor that was lifting off
it's plug. I replaced it and now it'll boot in a machine but
periodically the machine crashes. I'm considering replacing all 7 of
the non-solid caps and am seeking opinions.

Generally those between the 'fets and the load are the more
likely to fail,

I'm not sure which ones they are. I've uploaded some pics, mainly of the PCI
card to see if those things that I think are conductive polymer caps really
are.

The little ones on the PCI card are electrolytic not
polymer.



However, there's a pic of the ti4200's there, can you tell me which
caps are most likely to be the ones? I can only see what looks like one
power MOSFET and it's surrounded by caps.

I'd replace the following as pictured, if you don't have
enough spare caps on hand to do them all then I'd replace
them in waves by number, trying the card after each wave
(finishing one card before trying to start the second).

http://69.36.166.207/usr_1034/Bugs026-numbered.jpg

However, if none of them look even slightly bulged or leaky,
I'm less confident that replacing them will make a
difference.

The yellow one on the card on the left looks like someone
already repaired it, it's quite rare to see a manufacturer
do what they did putting a larger cap between one original
cap's positive pin and another's negative. I've done
repairs like that before but never saw it as an original cap
placement.



I'd replace those first if you have a few
spare capacitors lying around or that you could cannibalize
off of something else like a dead motherboard (though with
known good caps still).

I've always been a bit wary of re-using caps as I don't have a multimeter
cpable of testing capcitance and, IME, bad caps are the biggest reason for
failure for most electronics.

True they're one of the most likely causes but usually vent
out the top or bottom if they're bad enough to matter. I
wouldn't use some ancient socket 7 board for the donor caps
though, would tend to stick with Japanese well known brands.



There are five 1500uf 6.3V caps on each graphics card and I have a (dead)
Asus mobo with seven Nichicon caps on it of the same value. However, the
caps on the graphics cards are 8mm diameter with 3.5mm lead spacing and the
ones on the mobo are 10mm / 5mm and have very short leads. Maybe I can get
them to fit...

Maybe. If the leads are very short I doubt you'll be able
to make the 5mm spaced ones fit in the 3.5mm locations.
Look on the back of the card, sometimes manufacturers put
both lead spacings so they have the option of using either
cap size.

Don't worry so much about the values though, if the original
cap is rated for 16V be sure to use a replacement rated for
16V but with any others, use anything rated for 6V up to
around 16V and at least 1000uF, if worst came to worst you
might even try 470uF caps if they're 10V or 16V. Reusing
old parts, and having video cards that don't work, all
you're really losing out on is time if you don't want to buy
new caps by mail-order. Sometimes even given the best
effort there really was something else on the card that was
damaged and you have to give up, but having two cards and
one at least showing some signs of life, it seems fair odds
you can get at least one to work.

Something else you might try is taking the solid caps off
one if you can't get it to work and putting those on the
other one. It's much less likely the solid ones failed.


As for the smaller caps, I don't have exact matches in my box of bits
although I can get close (albeit with bigger components) but the caps are of
unknown quality.


If by the smaller ones you mean those with the turquiose
band, those are polymer and shouldn't need replaced. The
other smaller ones at the top of the picture were like the
one you said failed, right? If so they are definitely
suspect and again you don't necessarily need exact matches.

Generally a manufacturer doesn't use some value because it's
all that important, they're trying to strike a balance of
reliability and low cost. Just recognize that they seldom
use a 16V cap unless the circuit point is at least 12V,
unless they had used 16V caps in very many locations and had
just wanted to reduce the parts count. Otherwise everywhere
else with lower than 16V caps can use a 6V which is the most
common you'd find on many old motherboards.

Bigger components are fine, so long as you get the leads in
the holes with a good solder attachment and they're not
touching another part's leads, not so tall that you then
can't use an adjacent motherboard slot if you'd needed to.
A significantly larger sized higher ESR cap can sometimes be
used as a substitute for a smaller low-ESR cap in an
emergency, and in those cases it is generally better to use
a higher voltage value (within reason, moving to 100V is
probably unreasonable) than going with more than triple the
capacitance value unless the original cap was a lower-uF
polymer type then you may have more margin to use double, 6X
the value. It's hardly a science as we don't know the
original circuit parameters as built and if the cap doesn't
have a significantly lower ESR than the original then a
higher uF value may not cause significant enough addt'l
surge to matter.



Interestingly, it seems that Leadtek must have been going through a
rough patch when these cards were produced. Leadtek enjoy a large
market presence in New Zealand so consequently I have a few of their
cards around from over the years. The oldest is a PCI S3 Trio64V+
with 1MB RAM on-board. The newest is a 7800GT with 256MB RAM. I also
have a Leadtek GeForce 2 MX400, an FX5600 and an FX5700.

Here's the interesting bit. The old (and I mean 15 years old) PCI
card has all solid conductive polymer (CP) capacitors. The GF2 has
all ali can (AC) 'wet' caps but was made before the great capacitor
swindle at the end of last millenium and is still going strong in
daily usage.

Are you sure those are solid, conductive polymer capacitors
on the old card? In that era they tended to be mostly
tantalum (small rectangular yellow surface mounted) or small
electrolytics which may or may not have been surface mounted
without any vent slits so they did look like solid polymers
but often had a black stripe instead of colored to quickly
identify which they were.

I'm not sure, they just look like it. I've uploaded pics he

http://s308.photobucket.com/albums/kk349/misfitnz/


They're electrolytic, as is practically everything with a
black strip or that old in an aluminum can.



Both of the ti4200's have a mix of CP and AC caps but,
interestingly, the mix varies from one card to the other. (They were
bought 4 months apart, right after the public furore over the bad
caps scandal.) Also interestingly, one of them has two caps replaced
by one, bridging from the +ive of one cap stencil to the -ive of the
other. (There are two caps on the other PCB.)

All three of the later cards I have are fitted with all CP
capacitors.

Now, I'm thinking that, with the ti4200's, it might be worth
replacing the AC caps with new ones and hoping that the cards are
fixed. Correct me if I'm wrong but it's unlikely that any other
components would have failed causing the cards to fail. They've
always been in well ventilated cases and have never been
overclocked. Am I right? Silicon gates and resistors should still be
fine, chances are, if I replace the AC caps then the cards will work
and become reliable again?

If the cards were pulled out of the systems the moment they
ceased to stay stable there is a good chance they are ok
besides the capacitors, but it is still possible that other
parts were damaged because without the caps functioning
properly the card starts using more current and with ever
higher ripple, that can cause the regulation to go out of
control.

Both cards were pulled when they were failing but working and all other
components *appear* to be fine...

They seem like good candidates for recapping, it's just that
they're old enough at this point that the cost and time
weighs against it. Generally what I do when ordering caps
for some project is get extras, like a bunch of 8mm and 10mm
diameter in 10-12mm size and 20+mm size, then I have extras
for other projects so it's not so costly to order exactly
what I'd need each time.



I would try replacing the caps if it can be done cheaply or
at no expense.

I could probably do one card using caps I have lying around, salvaged from
other components. However, they're all an unknown quantity.

Quality, you can search for the brand and model at
badcaps.net, or maybe find their spec sheets. You can also
look at where they were in the old equipment you pulled them
out of, for example if they were in the CPU or memory
switching regulation circuits they are probably the right
type, and reasonably assumed still viable if they haven't
vented... and yet, the same could've been said about the
caps still on the video cards that don't look vented so you
can only try and see how it goes.



$20 seems a bit much unless prices
are a lot higher down there and you have to mail-order them.

Both actually. We're a small country and economies of scale come into it.
Also I have to mail-order or travel a fair distance to a store which may or
may not have the caps I require. :-(

You could always call them and ask, and see if they'd throw
a handful in an envelop for regular postal service delivery
to cut down on costs? Sometimes if you say you're a student
they might be more accomdating.


I'd see if I had an old motherboard lying around with
something close enough, that'll fit in the available space.

This might be my best option, try it with one of the graphics cards anyway.


That's what I usually do on anything that isn't of high
value anymore, then I started getting too big a stack of
cannibalized motherboards and started using a small torch to
pull lots of parts off and throw them into a few plastic
containers so I could throw the boards away.



Back when I first started replacing caps on mobos I re-used caps on a board
I really wanted to keep (a Gigabyte 440BX Socket 370, Coppermine-capable
board that had a Promise RAID controller on-board and would run at 133/33MHz
FSB/PCI) but I couldn't get it to run stably. I ended up replacing some caps
with salvaged caps a few times until I destroyed the board. That put me off
re-using salvaged caps.

Sometimes there's more than one problem. Failed caps plus
slight board or surface mount component cracks then handling
the board to fix the caps can make it even worse.
Motherboards in particular can be a real lottery, I've tried
to keep at least one backup board from every era just in
case I need a spare for myself or someone I'd sold a system
to. Probably half or more of them now have dead batteries.

My general rule of thumb on replacing with new caps is,
don't spend the money if a performance upgrade would matter
or if the cost exceeds half the value of the equipment, or
if the part wouldn't show any signs of life anymore if there
aren't any caps horribly deformed or vented, or if you don't
really need the part to work... for example I don't know the
market down there but I can generally pick up a used video
card from some shop or online from a hardware enthusiast
FS/FT forum for under $20 if they're not trying to gouge on
postage.




They're quite fully-featured cards, with monitoring chips on them
that tell Speedfan on-die temp, a thermistor between GPU and HS temp
and fan speed. They also have a series of three LEDs that tell if
they're running in 4x or 8x mode and if there is an AGP slot
problem. They've also got some pretty impressive lumps of aluminium
on both sides of the cards and the fans are still working and not
excessively noisy.

So... What do you think? It's not *so* much about the money really,
it's more about keeping them out of a landfill somewhere and also
justifying the outlay in, ummm, 2002 or so. I'd be quite bummed out
if i mail-ordered the caps, fitted them and still had a dead card
(or two, the postage on the caps means it'd be more sensible to buy
enough for both...)

Cheers, and TIA for any input.

I don't think it's worth the time, money, and uncertainty of
success to try it today. If I already had the caps lying
around or that could be cannibalized off a motherboard/etc
in a couple minutes time then I might try it... unless you
really need a couple AGP video cards and would end up
actually spending the $20+ on one if you don't get either of
them to work.

I don't *need* them per se, although it's always handy to have spares
around. I might tinker with one in a day or two using salvaged caps and see
how it goes.

Thanks for the input mate,


You might also pull the heatsinks off to be sure the
original thermal material hasn't cooked away over time.

In article , ~misfit~
writes

I've always been a bit wary of re-using caps as I don't have a multimeter
cpable of testing capcitance and, IME, bad caps are the biggest reason for
failure for most electronics.

You aren't just testing for capacitance, but need to test for ESR
(equivalent series resistance). Failing caps develop high ESR in
addition to reduced capacitance, causing them to heat up. This becomes
a vicious circle until the capacitor fails. The reduced capacitance
results in increased ripple and this is what cooks the MOSFETs and
coils.

It's not worth buying an ESR meter just to test a few caps. Just swap
out all the electrolytics. I wouldn't even bother with the PCI card -
that goes to landfill. The ti4200s are worth saving for the sake of
half an hour's work, and you'll have the satisfaction of a successful
fix.

There are five 1500uf 6.3V caps on each graphics card and I have a (dead)
Asus mobo with seven Nichicon caps on it of the same value. However, the
caps on the graphics cards are 8mm diameter with 3.5mm lead spacing and the
ones on the mobo are 10mm / 5mm and have very short leads. Maybe I can get
them to fit...

Why re-use caps from a motherboard? They are an unknown quantity, and
the caps in switching power supplies on mobos have a hard life.

Get the right caps new. The 10mm dia ones are 36 pence each on
http://rswww.com, part number 449-0851, and the 8mm dia ones are 24
pence each, p/n 526-0954. I'd expect prices to be similar in NZ. I
think RS has an Australian presence - worth checking.

Get the right caps new. The 10mm dia ones are 36 pence each on
http://rswww.com, part number 449-0851, and the 8mm dia ones are 24
pence each, p/n 526-0954. I'd expect prices to be similar in NZ. I
think RS has an Australian presence - worth checking.


RS Australia
http://australia.rs-online.com/web/

On Wed, 27 Aug 2008 11:51:08 +0100, Mike Tomlinson
wrote:

In article , ~misfit~
writes

I've always been a bit wary of re-using caps as I don't have a multimeter
cpable of testing capcitance and, IME, bad caps are the biggest reason for
failure for most electronics.

You aren't just testing for capacitance, but need to test for ESR
(equivalent series resistance). Failing caps develop high ESR in
addition to reduced capacitance, causing them to heat up. This becomes
a vicious circle until the capacitor fails. The reduced capacitance
results in increased ripple and this is what cooks the MOSFETs and
coils.

It's not worth buying an ESR meter just to test a few caps. Just swap
out all the electrolytics. I wouldn't even bother with the PCI card -
that goes to landfill. The ti4200s are worth saving for the sake of
half an hour's work, and you'll have the satisfaction of a successful
fix.

There are five 1500uf 6.3V caps on each graphics card and I have a (dead)
Asus mobo with seven Nichicon caps on it of the same value. However, the
caps on the graphics cards are 8mm diameter with 3.5mm lead spacing and the
ones on the mobo are 10mm / 5mm and have very short leads. Maybe I can get
them to fit...

Why re-use caps from a motherboard? They are an unknown quantity, and
the caps in switching power supplies on mobos have a hard life.

Get the right caps new. The 10mm dia ones are 36 pence each on
http://rswww.com, part number 449-0851, and the 8mm dia ones are 24
pence each, p/n 526-0954. I'd expect prices to be similar in NZ. I
think RS has an Australian presence - worth checking.

The reason to try reusing capacitors from a motherboard a

1) Capacitors from a motherboard don't generally fail below
the needed threshold for the video cards if they aren't
visibly vented.

2) Capacitors from a motherboard tend to be larger sized
lower ESR than the originals so even if they have had a
somewhat hard life, their parameters still tend to fall
within that required by the circuit after accounting for
aging, unless there were a defective formula that would
self-destruct before their expected lifespan had ended.

3) Buying new capacitors is easily worth the cost of one
entire video card as misfit has already priced them out at
$20 per card. That seems high to me but being in the US I
really can't say, I've heard of other areas having
significant addt'l expense sourcing other electronic
components so it wouldn't be surprising. By trying
motherboard caps it may be possible to determine if a
capacitor replacement will revive a card and if so, easier
to deem whether it is worth buying the capacitors. Further
it has been noted that another used card can be purchased
for about the same price, so while it would be rewarding to
get it working again the path towards one goal seems surer
by buying another card the owner claims does work.

4) I suggested it because I do it all the time with good
results. The one caveat is that this is almost always when
I'm replacing a cap that was visibly vented already, since
the others on these two cards have not vented it leaves some
uncertainty about the real cause since even a cap that isn't
functioning very well would have a signficantly lower
filtering requirement with the card just posting and
running in 2D mode long enough to see if the card worked at
all, versus operating in 3D mode like when gaming.

Also, if one has reasonable soldering skills and an
ohm-meter doesn't indicate a near short on any caps, the
original caps need not necessarily be removed from the card
at all, the proposed second cap can be placed on the opposte
side of the card and it's leads carefully soldered to the
exposed portion of the original cap's leads. Pretty it
isn't, but if the original simply isn't a suitably low ESR
or capacitance value anymore, the second will tend to be
enough to offset that plus it's usually quicker and easier
to just tack down a second cap on the rear of the card then
desolder it after testing the result.


Another thing ~misfit~ might try is fitting the PCI card in
a board, setting the board bios to consider PCI the primary
video, then installing the AGP card and seeing if the card
is detected by nvflash or equivalent. If it shows up it
might be worthwhile to try flashing a bios to it to see if
it had the EPROM contents corrupted and that is why one, not
the one that is just instable but rather the other one,
isn't functioning.

Please bear with me, this is getting long but I'm finding it useful. ;-)

I'd selectively snip but I want to save the post for future reference and it
will be easier all together.

Somewhere on teh intarweb "kony" typed:
On Wed, 27 Aug 2008 16:02:12 +1200, "~misfit~"
wrote:
Somewhere on teh intarweb "kony" typed:
On Wed, 27 Aug 2008 00:34:41 +1200, "~misfit~"
wrote:

Ok, so I have a couple Leadtek WinFast A280 LE AGP 8x VIVO 128MB
GeForce ti4200's. I bought them for my machine and my girlfriends
(back when I had one) and they cost over NZ$300 each new. They were
a significant investment at the time, costing twice as much as the
T'bredB Athlon XP CPUs they ran with.

My ex's card failed before we split and I put it on the shelf and
bought her a 7600GS to replace it. Mine failed not long afterwards
and also went on the shelf. I couldn't bring myself to throw them
away considering what they cost. Actually, only one is VIVO, the
other didn't need to be VIVO and was ~$40 cheaper.

The other day I was having one of my periodic clean-outs and I had
a good look at them. One of them had a capacitor that was lifting
off it's plug. I replaced it and now it'll boot in a machine but
periodically the machine crashes. I'm considering replacing all 7
of the non-solid caps and am seeking opinions.

Generally those between the 'fets and the load are the more
likely to fail,

I'm not sure which ones they are. I've uploaded some pics, mainly of
the PCI card to see if those things that I think are conductive
polymer caps really are.

The little ones on the PCI card are electrolytic not
polymer.

OK, thanks. The only caps I've noticed that look like those before are
polymer. Usually IME 'lytics look more like... well, 'lytics really.

However, there's a pic of the ti4200's there, can you tell me which
caps are most likely to be the ones? I can only see what looks like
one power MOSFET and it's surrounded by caps.

I'd replace the following as pictured, if you don't have
enough spare caps on hand to do them all then I'd replace
them in waves by number, trying the card after each wave
(finishing one card before trying to start the second).

http://69.36.166.207/usr_1034/Bugs026-numbered.jpg

However, if none of them look even slightly bulged or leaky,
I'm less confident that replacing them will make a
difference.

Ok. Looking at the card on the right on your pic, the cap that has an arrow
pointing to it situated near the connectors is one I soldered in. The
original, same specs and brand as the cap you've numbered 1 (330uf, 6.3V),
was the only cap on either board that showed any visible signs of
deterioration. It had lifted off it's plug to the extent that I'm surprised
the plug is still holding in there.

As that cap failed and the cap that you've put as number one to replace were
identical it wouldn't surprise me if, although visually fine, the other (#1)
is way out of spec.

They're GSC brand, as are all the other 'lytics, a name I've seen lots of
times on caps I've had to replace. The replacement I fitted is a 25V Jenpo.
I went with matching the capacitance as long as the voltage was the same or
higher and it's all I found in my small collection of caps. I've only
bothered keeping caps if either they came out of quality equipment (IBM
servers for instance that were 'retired' due to obsolescence, nothing wrong
with them) or if I recognise the brand of cap as being a good one such as
Nichicon or Rubycon.

The yellow one on the card on the left looks like someone
already repaired it, it's quite rare to see a manufacturer
do what they did putting a larger cap between one original
cap's positive pin and another's negative. I've done
repairs like that before but never saw it as an original cap
placement.

I bought that card like that, as a new graphics card in a sealed anti-static
bag inside a sealed box. I didn't notice at the time, I was just keen on
getting it in my machine to replace the GF 2 MX that I had been using up
until then. It wasn't until recently that I even noticed it. I've just
examined the other side of the board and it looks like there has never been
anything soldered through the other holes so I'd say it came out of the
factory like that. Looking at the back of the board they don't appear to be
wired in parallel but I suppose they must be. The card worked fine like that
for 2 years.

That yellow cap is a 680uf/.6.3V component. The two in the same place on the
other card are 560uf/4V (polymer) and 330uf/6.3V ('lytic).

I'd replace those first if you have a few
spare capacitors lying around or that you could cannibalize
off of something else like a dead motherboard (though with
known good caps still).

I've always been a bit wary of re-using caps as I don't have a
multimeter cpable of testing capcitance and, IME, bad caps are the
biggest reason for failure for most electronics.

True they're one of the most likely causes but usually vent
out the top or bottom if they're bad enough to matter.

I've replaced seemingly perfect caps on non-working mobos (in the VRM area)
before and had them work perfectly afterwards.

I
wouldn't use some ancient socket 7 board for the donor caps
though, would tend to stick with Japanese well known brands.

No. As I said above, I only keep caps that I think have a good chance of
being in spec.

There are five 1500uf 6.3V caps on each graphics card and I have a
(dead) Asus mobo with seven Nichicon caps on it of the same value.
However, the caps on the graphics cards are 8mm diameter with 3.5mm
lead spacing and the ones on the mobo are 10mm / 5mm and have very
short leads. Maybe I can get them to fit...

Maybe. If the leads are very short I doubt you'll be able
to make the 5mm spaced ones fit in the 3.5mm locations.

Yeah, I've had that problem more times than I care to remember.

Look on the back of the card, sometimes manufacturers put
both lead spacings so they have the option of using either
cap size.

LOL, already did, no luck. The irony is, the Asus mobo with the Nichicon
caps that I'll probably use as a donor? It has double-spacings to take 8mm
or 10mm caps and is fitted with the 10mm ones.

Actually I've just found a couple VRM modules out of an IBM Netfinity that
have 10mm/1500/6.3 caps on them but appear to be hand-made. There are longer
leads, bent over on the back. I might unsolder those as the leads are
longer.

The single biggest problem I've had, that I often have with replacing caps,
is getting them small enough to fit. Just about all replacements I see for
sale are physically larger.

Don't worry so much about the values though, if the original
cap is rated for 16V be sure to use a replacement rated for
16V but with any others, use anything rated for 6V up to
around 16V and at least 1000uF, if worst came to worst you
might even try 470uF caps if they're 10V or 16V. Reusing
old parts, and having video cards that don't work, all
you're really losing out on is time if you don't want to buy
new caps by mail-order. Sometimes even given the best
effort there really was something else on the card that was
damaged and you have to give up, but having two cards and
one at least showing some signs of life, it seems fair odds
you can get at least one to work.

I hope so. Both cards will boot but with lines through the BIOS screen (then
I turn the PC off). Interestingly, they both show *exactly* the same thing,
groups of three vertical green lines in six places across the screen in
exactly the same place. Actually, that's a worry.

Something else you might try is taking the solid caps off
one if you can't get it to work and putting those on the
other one. It's much less likely the solid ones failed.


Yeah. I'll try to get the VIVO card going first, even though it's the one
with the bridged cap.

As for the smaller caps, I don't have exact matches in my box of bits
although I can get close (albeit with bigger components) but the
caps are of unknown quality.


If by the smaller ones you mean those with the turquiose
band, those are polymer and shouldn't need replaced.

OK, I didn't realise that. Isuppose the fact that they're rated at 4V should
have been a giveaway. Most lytics I see are 6.3V. They're Sanyo too.

The
other smaller ones at the top of the picture were like the
one you said failed, right? If so they are definitely
suspect and again you don't necessarily need exact matches.

Yep. As I said above (I should have read down a bit). I'll try to find
suitable replacements. It's hard when they're so small though. I'll find
some caps somewhere.

Generally a manufacturer doesn't use some value because it's
all that important, they're trying to strike a balance of
reliability and low cost. Just recognize that they seldom
use a 16V cap unless the circuit point is at least 12V,
unless they had used 16V caps in very many locations and had
just wanted to reduce the parts count. Otherwise everywhere
else with lower than 16V caps can use a 6V which is the most
common you'd find on many old motherboards.

Bigger components are fine, so long as you get the leads in
the holes with a good solder attachment and they're not
touching another part's leads, not so tall that you then
can't use an adjacent motherboard slot if you'd needed to.
A significantly larger sized higher ESR cap can sometimes be
used as a substitute for a smaller low-ESR cap in an
emergency, and in those cases it is generally better to use
a higher voltage value (within reason, moving to 100V is
probably unreasonable) than going with more than triple the
capacitance value unless the original cap was a lower-uF
polymer type then you may have more margin to use double, 6X
the value. It's hardly a science as we don't know the
original circuit parameters as built and if the cap doesn't
have a significantly lower ESR than the original then a
higher uF value may not cause significant enough addt'l
surge to matter.

All good info, thanks.

Interestingly, it seems that Leadtek must have been going through a
rough patch when these cards were produced. Leadtek enjoy a large
market presence in New Zealand so consequently I have a few of
their cards around from over the years. The oldest is a PCI S3
Trio64V+ with 1MB RAM on-board. The newest is a 7800GT with 256MB
RAM. I also have a Leadtek GeForce 2 MX400, an FX5600 and an
FX5700.

Here's the interesting bit. The old (and I mean 15 years old) PCI
card has all solid conductive polymer (CP) capacitors. The GF2 has
all ali can (AC) 'wet' caps but was made before the great capacitor
swindle at the end of last millenium and is still going strong in
daily usage.

Are you sure those are solid, conductive polymer capacitors
on the old card? In that era they tended to be mostly
tantalum (small rectangular yellow surface mounted) or small
electrolytics which may or may not have been surface mounted
without any vent slits so they did look like solid polymers
but often had a black stripe instead of colored to quickly
identify which they were.

I'm not sure, they just look like it. I've uploaded pics he

http://s308.photobucket.com/albums/kk349/misfitnz/

They're electrolytic, as is practically everything with a
black strip or that old in an aluminum can.

Ok.

Both of the ti4200's have a mix of CP and AC caps but,
interestingly, the mix varies from one card to the other. (They
were bought 4 months apart, right after the public furore over the
bad caps scandal.) Also interestingly, one of them has two caps
replaced by one, bridging from the +ive of one cap stencil to the
-ive of the other. (There are two caps on the other PCB.)

All three of the later cards I have are fitted with all CP
capacitors.

Now, I'm thinking that, with the ti4200's, it might be worth
replacing the AC caps with new ones and hoping that the cards are
fixed. Correct me if I'm wrong but it's unlikely that any other
components would have failed causing the cards to fail. They've
always been in well ventilated cases and have never been
overclocked. Am I right? Silicon gates and resistors should still
be fine, chances are, if I replace the AC caps then the cards will
work and become reliable again?

If the cards were pulled out of the systems the moment they
ceased to stay stable there is a good chance they are ok
besides the capacitors, but it is still possible that other
parts were damaged because without the caps functioning
properly the card starts using more current and with ever
higher ripple, that can cause the regulation to go out of
control.

Both cards were pulled when they were failing but working and all
other components *appear* to be fine...

They seem like good candidates for recapping, it's just that
they're old enough at this point that the cost and time
weighs against it.

Yeah. Time isn't an issue for me as I'm an invalid and have nothing but
time. That makes the cost an issue though. :-(

Generally what I do when ordering caps
for some project is get extras, like a bunch of 8mm and 10mm
diameter in 10-12mm size and 20+mm size, then I have extras
for other projects so it's not so costly to order exactly
what I'd need each time.

Yeah, I've done that to a certain extent. It's really hard to get caps here
though. It seems that all manufacturers source their components offshore
which leaves hobbyists high and dry as far as locally sourced components go.
I was after some caps a couple years ago to repair a mobo and simply
couldn't get any suitable anywhere I looked here. I contacted the NZ Rubycon
agent and asked. They said they'd get back to me by email. I got an email a
week later saying that they didn't have stock in NZ and there'd be a 2 month
lead-time for stock ex-Japan. I said OK, I want some. They replied a week
later saying it'd be a minimum order of 1,000 units payable in advance. I
said, OK, how much would that cost me? I never heard back, despite
re-sending the email weekly for a month.

In the end I got a handful from Badcaps.net, sent to a friend in the US as
they wouldn't post to here and he sent them on to me. Bloody expensive.

I would try replacing the caps if it can be done cheaply or
at no expense.

I could probably do one card using caps I have lying around,
salvaged from other components. However, they're all an unknown
quantity.

Quality, you can search for the brand and model at
badcaps.net, or maybe find their spec sheets. You can also
look at where they were in the old equipment you pulled them
out of, for example if they were in the CPU or memory
switching regulation circuits they are probably the right
type, and reasonably assumed still viable if they haven't
vented... and yet, the same could've been said about the
caps still on the video cards that don't look vented so you
can only try and see how it goes.

Cheers.

$20 seems a bit much unless prices
are a lot higher down there and you have to mail-order them.

Both actually. We're a small country and economies of scale come
into it. Also I have to mail-order or travel a fair distance to a
store which may or may not have the caps I require. :-(

You could always call them and ask, and see if they'd throw
a handful in an envelop for regular postal service delivery
to cut down on costs? Sometimes if you say you're a student
they might be more accomdating.

It seems that even the places that cater to hobbyists are getting gouged.
Anywhere that stocks them seem to charge a minimum of a buck a cap, going up
to 3 bucks and have a minimum frieght price. That's why I enquired about
getting a batch from Rubycon. I figured I could either re-cap some boards to
make my money back or advertise some of the surplus in the local comp
group..

I'd see if I had an old motherboard lying around with
something close enough, that'll fit in the available space.

This might be my best option, try it with one of the graphics cards
anyway.

That's what I usually do on anything that isn't of high
value anymore, then I started getting too big a stack of
cannibalized motherboards and started using a small torch to
pull lots of parts off and throw them into a few plastic
containers so I could throw the boards away.

Heh! I had a big cleanout a while back, kept a stack of boards to
cannabalise but got sick of it after a short time. I don't have a torch and
it was tedious removing them with my 30W soldering iron.

Back when I first started replacing caps on mobos I re-used caps on
a board I really wanted to keep (a Gigabyte 440BX Socket 370,
Coppermine-capable board that had a Promise RAID controller on-board
and would run at 133/33MHz FSB/PCI) but I couldn't get it to run
stably. I ended up replacing some caps with salvaged caps a few
times until I destroyed the board. That put me off re-using salvaged
caps.

Sometimes there's more than one problem. Failed caps plus
slight board or surface mount component cracks then handling
the board to fix the caps can make it even worse.
Motherboards in particular can be a real lottery, I've tried
to keep at least one backup board from every era just in
case I need a spare for myself or someone I'd sold a system
to. Probably half or more of them now have dead batteries.

I keep a few spares too. I'm not "in the business" but I've sold a few
second-hand systems and built a few new systems to order for folks. I try to
keep a few spares on-hand. I guess that's part of the motivation behind
trying to fix these graphics cards. Handy to have a reasonably-powerful
spare card or two on hand.

My general rule of thumb on replacing with new caps is,
don't spend the money if a performance upgrade would matter
or if the cost exceeds half the value of the equipment, or
if the part wouldn't show any signs of life anymore if there
aren't any caps horribly deformed or vented, or if you don't
really need the part to work... for example I don't know the
market down there but I can generally pick up a used video
card from some shop or online from a hardware enthusiast
FS/FT forum for under $20 if they're not trying to gouge on
postage.

I just paid NZ$45 on the local version of ebay (trademe.co.nz) for an FX5700
in good condition, all polymer caps and with original box, manual, driver
CDs etc. GF2's still regularly fetch $25 each. You can get cheaper if you're
prepared to buy items where they say "untested, working when stored, no
warranty" but I know of people who sell duds that way and can't afford to
gamble.

They're quite fully-featured cards, with monitoring chips on them
that tell Speedfan on-die temp, a thermistor between GPU and HS
temp and fan speed. They also have a series of three LEDs that
tell if they're running in 4x or 8x mode and if there is an AGP
slot problem. They've also got some pretty impressive lumps of
aluminium on both sides of the cards and the fans are still
working and not excessively noisy.

So... What do you think? It's not *so* much about the money really,
it's more about keeping them out of a landfill somewhere and also
justifying the outlay in, ummm, 2002 or so. I'd be quite bummed out
if i mail-ordered the caps, fitted them and still had a dead card
(or two, the postage on the caps means it'd be more sensible to buy
enough for both...)

Cheers, and TIA for any input.

I don't think it's worth the time, money, and uncertainty of
success to try it today. If I already had the caps lying
around or that could be cannibalized off a motherboard/etc
in a couple minutes time then I might try it... unless you
really need a couple AGP video cards and would end up
actually spending the $20+ on one if you don't get either of
them to work.

I don't *need* them per se, although it's always handy to have spares
around. I might tinker with one in a day or two using salvaged caps
and see how it goes.

Thanks for the input mate,

You might also pull the heatsinks off to be sure the
original thermal material hasn't cooked away over time.

Already done. :-)

Cheers,
--
Shaun.

DISCLAIMER: If you find a posting or message from me
offensive, inappropriate, or disruptive, please ignore it.
If you don't know how to ignore a posting, complain to
me and I will be only too happy to demonstrate... ;-)

Somewhere on teh intarweb "kony" typed:
On Wed, 27 Aug 2008 18:30:50 +1200, "~misfit~"
wrote:

Somewhere on teh intarweb "Phil Weldon" typed:
'~misfit~' wrote, in part:
I could probably do one card using caps I have lying around,
salvaged from other components. However, they're all an unknown
quantity.
_____

Write down the identification letters and numerals on a capacitor.
Enter this as a search string in Google. You will likely get some
hits that will identify the capacitor type, maximum working voltage,
and capacitance.
Only the electrolytic high capacitance (200 micro-farad or more)
units in a switching voltage converter/regulator are likely to have
gone bad. The proper replacements must be low ESR (Equivalent
Series Resistance) parts. If the ESR is too high for the purpose an
electrolytic capacitor will overheat, the electrolyte will expand,
and the top will pop (or at the least the internal structure will
rupture and the the capacitor will open or short.)

Thanks Phil. I've done a few capacitor replacements on motherboards
and always get Low ESR parts. I've always thought that there was
really only three other things that I needed to know to match them
up (besides size) being capacitance, voltage and temp rating.

True, but those don't necessarily need to match.
On any decent low-ESR cap it will be 105C temp rated or
higher, that's almost automatic.

Voltage, generally you have a tiny penalty going to
significantly higher voltage than needed, but practically
speaking it doesn't matter, rather what size it is would.

Keep in mind that often when an electrolytic is used, they
use a certain uF value not because it needs that, they use
that uF value because a smaller uF value cap has higher ESR
(assuming still within the same family of capacitor). This
can often be seen when a manufacturer has substituted a
solid polymer for an electrolytic or vice-versa, that the
electrolytic is always a higher uF value if they had any
hope of it remaining a functional circuit. So basically,
with a low ESR electrolytic cap that is rated for 16V or
less, you could get away with trying anything small enough
to fit, the larger the better since you wouldn't be able to
get one to fit if it were so large to create excessive
inrush current.

For example if one were a 6V, 2200uF value, you could
probably get away with using a 16V, 1000uF, a 10V 1500uF, or
higher uF per voltage. Depends on the cap, and design since
some caps may be in parallel. As for voltage, if you have
an old board lying around (with an AGP slot of course) put
the card in, turn the system on, and measure the voltage
across the cap leads. Most will be below 6V so 6V caps are
a practial value to use. Some might be at 12V but

When I said "unknown quantity" above I meant that I don't know if
they're still within spec as they're salvaged components taken from
used, usually non-working circuits. I mean, I know that, if they're
Rubycon, Nichicon, Sanyo or Teapo that they're likely to be fine but
other than that I have no way of testing them

Usually caps do vent if they're a failure point in the
circuit. At worst you make your best guess about which
replacements are ok, solder them in, probe their leads for
continuity to be sure there isn't a short (observing
polarity, meter will beep for a moment as the cap charges),
then put in a non-valuable motherboard to test the result.

Thanks for the info, all good. :-)

Cheers,
--
Shaun.

DISCLAIMER: If you find a posting or message from me
offensive, inappropriate, or disruptive, please ignore it.
If you don't know how to ignore a posting, complain to
me and I will be only too happy to demonstrate... ;-)

On Thu, 28 Aug 2008 00:32:18 +1200, "~misfit~"
wrote:


They're GSC brand, as are all the other 'lytics, a name I've seen lots of
times on caps I've had to replace. The replacement I fitted is a 25V Jenpo.
I went with matching the capacitance as long as the voltage was the same or
higher and it's all I found in my small collection of caps. I've only
bothered keeping caps if either they came out of quality equipment (IBM
servers for instance that were 'retired' due to obsolescence, nothing wrong
with them) or if I recognise the brand of cap as being a good one such as
Nichicon or Rubycon.

Being GSC, I'd expect to have to replace all of those for
longer term use.


Generally what I do when ordering caps
for some project is get extras, like a bunch of 8mm and 10mm
diameter in 10-12mm size and 20+mm size, then I have extras
for other projects so it's not so costly to order exactly
what I'd need each time.

Yeah, I've done that to a certain extent. It's really hard to get caps here
though. It seems that all manufacturers source their components offshore
which leaves hobbyists high and dry as far as locally sourced components go.
I was after some caps a couple years ago to repair a mobo and simply
couldn't get any suitable anywhere I looked here. I contacted the NZ Rubycon
agent and asked. They said they'd get back to me by email. I got an email a
week later saying that they didn't have stock in NZ and there'd be a 2 month
lead-time for stock ex-Japan. I said OK, I want some. They replied a week
later saying it'd be a minimum order of 1,000 units payable in advance. I
said, OK, how much would that cost me? I never heard back, despite
re-sending the email weekly for a month.

In the end I got a handful from Badcaps.net, sent to a friend in the US as
they wouldn't post to here and he sent them on to me. Bloody expensive.

I suppose it depends on how many you need, how you need them
shipped, but I thought I vaguely recalled our US Postal
Service will deliver a large envelop for a little under $10
USD. A better (cheaper) place to get some other good brands
(just not Rubycon) might be electronics houses like Digikey
or Mouser. Depending on what the cap size is, sometimes
they're still around a dollar a piece (in Qty. 10+) for the
larger ones I've sometimes used in power supplies, though
probably closer to $0.25 each per 10 in the smaller sizes
you're dealing with.


You could always call them and ask, and see if they'd throw
a handful in an envelop for regular postal service delivery
to cut down on costs? Sometimes if you say you're a student
they might be more accomdating.

It seems that even the places that cater to hobbyists are getting gouged.
Anywhere that stocks them seem to charge a minimum of a buck a cap, going up
to 3 bucks and have a minimum frieght price. That's why I enquired about
getting a batch from Rubycon. I figured I could either re-cap some boards to
make my money back or advertise some of the surplus in the local comp
group..

I'd see if I had an old motherboard lying around with
something close enough, that'll fit in the available space.

This might be my best option, try it with one of the graphics cards
anyway.

That's what I usually do on anything that isn't of high
value anymore, then I started getting too big a stack of
cannibalized motherboards and started using a small torch to
pull lots of parts off and throw them into a few plastic
containers so I could throw the boards away.

Heh! I had a big cleanout a while back, kept a stack of boards to
cannabalise but got sick of it after a short time. I don't have a torch and
it was tedious removing them with my 30W soldering iron.

I can imagine it would be. I've used an iron for something
now and then that was still on a board but usually just heat
a big area with a torch and slap the side of a board against
a piece of plywood and a bunch of stuff falls off. I don't
care much if I get everything because I've had a source of
dead boards from some shops so it's really more about
putting the least amount of work possible into doing it.
Plus it really sticks when the board is heated by a torch,
has to be done outside.


I just paid NZ$45 on the local version of ebay (trademe.co.nz) for an FX5700
in good condition, all polymer caps and with original box, manual, driver
CDs etc. GF2's still regularly fetch $25 each. You can get cheaper if you're
prepared to buy items where they say "untested, working when stored, no
warranty" but I know of people who sell duds that way and can't afford to
gamble.

I have to keep reminding myself that converts to less, about
1:0.7 IIRC, in US:NZ dollars... so the latter would only be
$17.50 USD which is not so bad, almost getting to the point
where the shipping cost approaches half the card cost.

Somewhere on teh intarweb "Mike Tomlinson" typed:
In article , ~misfit~
writes

I've always been a bit wary of re-using caps as I don't have a
multimeter cpable of testing capcitance and, IME, bad caps are the
biggest reason for failure for most electronics.

You aren't just testing for capacitance, but need to test for ESR
(equivalent series resistance). Failing caps develop high ESR in
addition to reduced capacitance, causing them to heat up. This
becomes a vicious circle until the capacitor fails. The reduced
capacitance results in increased ripple and this is what cooks the
MOSFETs and coils.

It's not worth buying an ESR meter just to test a few caps. Just swap
out all the electrolytics. I wouldn't even bother with the PCI card -
that goes to landfill. The ti4200s are worth saving for the sake of
half an hour's work, and you'll have the satisfaction of a successful
fix.

There are five 1500uf 6.3V caps on each graphics card and I have a
(dead) Asus mobo with seven Nichicon caps on it of the same value.
However, the caps on the graphics cards are 8mm diameter with 3.5mm
lead spacing and the ones on the mobo are 10mm / 5mm and have very
short leads. Maybe I can get them to fit...

Why re-use caps from a motherboard? They are an unknown quantity, and
the caps in switching power supplies on mobos have a hard life.

Get the right caps new. The 10mm dia ones are 36 pence each on
http://rswww.com, part number 449-0851, and the 8mm dia ones are 24
pence each, p/n 526-0954. I'd expect prices to be similar in NZ. I
think RS has an Australian presence - worth checking.

Ok, thanks. I have the NZ rs-online bookmarked:

http://newzealand.rs-online.com/web/

Also I actually have their new (April '08 - March '09) ~3,000 page printed
catalogue next to my desk. I've been registered with them for years but have
found it seems that I have too little knowledge to find the parts I need.
Hence, although I get their emails regularly and a couple weeks ago
requested their catalogue (hoping it'd be easier to find the parts I need)
I've never ordered from them.

The thing that's been confusing me is that I've been unable to find "Low
ESR" caps and both of those you mention don't say that they're low ESR. How
am I supposed to know?

In their catalogue they list various electrolytics as being suitable for
hifi crossovers and other various uses but none of their catgories say "low
ESR" or "suitable for computer components".

Oh, NZ$1.46 and $1.09 each respectively in lots of 5+. (About right
considering the exchange rate.) Fright (LOL, typo for freight, quite apt) is
a whole other ballgame. I really need to order a large bunch to make it
viable. However, I also need to know that what I'm ordering are the correct
parts. As mentioned previously, I'm poverty-stricken (otherwise I wouldn't
be tinkering with these old cards) and can't afford mistakes.

Cheers,
--
Shaun.

DISCLAIMER: If you find a posting or message from me
offensive, inappropriate, or disruptive, please ignore it.
If you don't know how to ignore a posting, complain to
me and I will be only too happy to demonstrate... ;-)

Somewhere on teh intarweb "Robert McMillan" typed:
Get the right caps new. The 10mm dia ones are 36 pence each on
http://rswww.com, part number 449-0851, and the 8mm dia ones are 24
pence each, p/n 526-0954. I'd expect prices to be similar in NZ. I
think RS has an Australian presence - worth checking.


RS Australia
http://australia.rs-online.com/web/

Thanks Robert, I have the NZ site bookmarked. See my reply to Mike.

Cheers,
--
Shaun.

DISCLAIMER: If you find a posting or message from me
offensive, inappropriate, or disruptive, please ignore it.
If you don't know how to ignore a posting, complain to
me and I will be only too happy to demonstrate... ;-)

On Fri, 29 Aug 2008 12:20:05 +1200, "~misfit~"
wrote:


The thing that's been confusing me is that I've been unable to find "Low
ESR" caps and both of those you mention don't say that they're low ESR. How
am I supposed to know?



Usually, low-esr is a prominent feature of a cap and will be
mentioned. If a daft seller doesn't describe it well enough
you are left going to the cap manufacturer's website and
consulting the spec sheet. If the manufacturer doesn't have
an online presence or spec sheet, or if that doesn't mention
low-ESR, you may assume it isn't.