Singing Grapics Card????

[~misfit~]

I have an HD7770 graphics card in my desktop. Just a basic one but a big
step up from the HD5670 I was using before for light gaming.

Last week I saw an auction for a faulty one on TM
http://www.trademe.co.nz/Browse/Listing.aspx?id=803566559 and put in an
autobid of $6.66 and won it. I thought that, at the very least I'd have a
really good direct-contact heatpipe cooler that I might be able to graft
onto my card (which has a simple ali block HS). Also though, being the
eternal optomist (despite life proving that it's an unrealistic outlook) I
hoped that I might be able to find a fault with it and fix it.

It arrived today and I fitted it to my spare / guest machine in place of the
HD5670 that was in there. Much to my delight it displayed the BIOS screen
just fine then booted into Win7 (albeit at a low resolution). A Windows
popup appeared saying 'installing drivers' and about that time I noticed a
sound that was akin to a very small mouse getting a filling. It was very
quiet and I might have missed it if I'd fitted the cover back on the PC....

The machine installed the new drivers fine and asked to reboot - I clicked
OK. It started up at full resolution! However the tiny mouse was still
getting dental work done - maybe moreso. I started Path of Exile and it
downloaded updates (I hadn't used the machine for a while) and then the load
screen came up - and the monitor went blank. :-(

However the fact that it ran for ~10 minutes makes me hopeful. I can't
understand why the owner didn't get a replacement under warranty - it's an
ASUS after all and the card isn't *that* old. <whirr click clang.> Oh,
unless they did and I now have the faulty one. :-/

I did a dirty shut-down of the machine then came to my laptop to post this
and ask for advice. Am I right in thinking that it may be an inductor (or
possibly capacitor) that's faulty? I can likely replace either if it is and
I can get a suitable replacement but I don't want to go blindly replacing
components. I was unable to pin-point the sound because of the large shroud
and heatsink / radiator covering the PCB. Frankly I'm surprised that I heard
it as I'm sure I have age-related hearing loss of higher frequencies.

What do you folks think? Any relevant advice? I've replaced mobo caps before
and even done a bit of surface-mount reworking - about as much as you can do
without a hot-air station at least. I'd be over the moon if I could repair
it - it'd likely be the only Xmas present I get.

Cheers,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[Paul]

I have an HD7770 graphics card in my desktop. Just a basic one but a big
step up from the HD5670 I was using before for light gaming.

Last week I saw an auction for a faulty one on TM
http://www.trademe.co.nz/Browse/Listing.aspx?id=803566559 and put in an
autobid of $6.66 and won it. I thought that, at the very least I'd have a
really good direct-contact heatpipe cooler that I might be able to graft
onto my card (which has a simple ali block HS). Also though, being the
eternal optomist (despite life proving that it's an unrealistic outlook) I
hoped that I might be able to find a fault with it and fix it.

It arrived today and I fitted it to my spare / guest machine in place of the
HD5670 that was in there. Much to my delight it displayed the BIOS screen
just fine then booted into Win7 (albeit at a low resolution). A Windows
popup appeared saying 'installing drivers' and about that time I noticed a
sound that was akin to a very small mouse getting a filling. It was very
quiet and I might have missed it if I'd fitted the cover back on the PC....

The machine installed the new drivers fine and asked to reboot - I clicked
OK. It started up at full resolution! However the tiny mouse was still
getting dental work done - maybe moreso. I started Path of Exile and it
downloaded updates (I hadn't used the machine for a while) and then the load
screen came up - and the monitor went blank. :-(

However the fact that it ran for ~10 minutes makes me hopeful. I can't
understand why the owner didn't get a replacement under warranty - it's an
ASUS after all and the card isn't *that* old. <whirr click clang.> Oh,
unless they did and I now have the faulty one. :-/

I did a dirty shut-down of the machine then came to my laptop to post this
and ask for advice. Am I right in thinking that it may be an inductor (or
possibly capacitor) that's faulty? I can likely replace either if it is and
I can get a suitable replacement but I don't want to go blindly replacing
components. I was unable to pin-point the sound because of the large shroud
and heatsink / radiator covering the PCB. Frankly I'm surprised that I heard
it as I'm sure I have age-related hearing loss of higher frequencies.

What do you folks think? Any relevant advice? I've replaced mobo caps before
and even done a bit of surface-mount reworking - about as much as you can do
without a hot-air station at least. I'd be over the moon if I could repair
it - it'd likely be the only Xmas present I get.

Cheers,

You have better be a switching power converter expert. The video card
needs to make a 1 volt supply at high current, to run the core of the
GPU chip.

Tbe easy problems, are leaking caps, because you can see them. Video cards
(the ones over $50), were the first electronics to make use of polymer
caps, so you'd hope the traditional leaky cap problem wouldn't be present.
I expect all three of my FX5200 cards, those use regular electrolytic caps.

A leaking cap can fail enough, to cause an inductor (toroid, powered
iron core), to burn up. Or a MOSFET could fail. Some power converter
circuits, measure fault current, by using the channel resistance
of the MOSFET as a means to measure current flow. Some of the
high end cards, use Volterra designs, and the principles of those
are different enough, it probably doesn't match your symptoms.
(Nothing on a Volterra, would be in audible range.)

Even if I had a decent scope handy, I doubt I could understand
what was wrong with one. Sure, a stuck-at fault, a short to
ground, I'd get that. But not a slight parametric problem.
I wouldn't know what was "normal".

*******

If you're absolutely determined to get $7 worth of fun out of
it. remove the heatsink, find the regulator IC, get the part number
off it, and download the PDF datasheet. That'll give you something
technical to work with. If you're lucky, the regulator is on
the solder side, and might be visible without any work.

OK, in the first picture, the core supply is at least three phase,
and uses small dual MOSFETs in 8 pin packages (cheep). And I cannot
locate a regulator chip! That's on the left hand side of the picture.
Some lower current regulators are located on the right hand side,
and those should be for memory voltages. Maybe one regulator on the
right is for VCC and one for VTT (terminator with push/pull capability).

http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/60_xfx777be_pcb_big.jpg

And there's nothing all that interesting on the secondary side (solder
side). The big chips there, could be digital format conversion,
going from proprietary GPU video out bus, to dual link DVI on a
connector. Something like that. Or maybe it's a VGA output circuit.

http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/56_xfx777be_frr_big.jpg

The thing uses Polymer caps (no stress relief lines on top).
So it's not likely to be a leaky cap problem. The odds of you
spotting a visual problem, are just about zero. If the problem
was going to be that easy to diagnose, it would probably
be smoking by now.

Good luck,
Paul

 

[~misfit~]

Thanks for replying Paul, I appreciate someone of your expertise giving
input.

You have better be a switching power converter expert.

I'm not an expert on anything - oh, I guess I qualify as an expert on the
management of semi-closed aquatic systems. I used to get paid around $100 /
hour for designing filtration systems for public aquaria a couple of decades
ago (when that was *really* good money).

The video card
needs to make a 1 volt supply at high current, to run the core of the
GPU chip.

Close enough. 1.1v when working and 0.9v idling - maybe 80 amps or so.

Tbe easy problems, are leaking caps, because you can see them. Video
cards (the ones over $50), were the first electronics to make use of
polymer caps, so you'd hope the traditional leaky cap problem
wouldn't be present. I expect all three of my FX5200 cards, those use
regular electrolytic caps.

Yeah, back in the day I used to replace leaky cheap caps on otherwise fine
motherboards with good quality low ESR Rubycons or similar. Some of the
boards that I re-capped are still running 10+ years later. The last couple
of builds I did I made sure that the mobos had all-solid polymer capacitors
though. I just came across a bare FX5200 board in my junk box when looking
for a suitable cap and it has wet 'lytics. Surface-mount but with stress
relief stampings on the tops.

A leaking cap can fail enough, to cause an inductor (toroid, powered
iron core), to burn up. Or a MOSFET could fail. Some power converter
circuits, measure fault current, by using the channel resistance
of the MOSFET as a means to measure current flow. Some of the
high end cards, use Volterra designs, and the principles of those
are different enough, it probably doesn't match your symptoms.
(Nothing on a Volterra, would be in audible range.)

An electronics whizz who I used to know told me that almost always when an
eltrical device makes noise it's vibrations in inductor coils and that's why
the better high frequency ones are epoxy potted. That's why I suspected it
was an inductor - though *why* it's singing is the question I guess....

Even if I had a decent scope handy, I doubt I could understand
what was wrong with one. Sure, a stuck-at fault, a short to
ground, I'd get that. But not a slight parametric problem.
I wouldn't know what was "normal".

*******

If you're absolutely determined to get $7 worth of fun out of
it. remove the heatsink, find the regulator IC, get the part number
off it, and download the PDF datasheet. That'll give you something
technical to work with. If you're lucky, the regulator is on
the solder side, and might be visible without any work.

I don't mind some 'work'. I'm a very poor invalid with nothing but time.

OK, in the first picture, the core supply is at least three phase,
and uses small dual MOSFETs in 8 pin packages (cheep). And I cannot
locate a regulator chip! That's on the left hand side of the picture.
Some lower current regulators are located on the right hand side,
and those should be for memory voltages. Maybe one regulator on the
right is for VCC and one for VTT (terminator with push/pull
capability).
http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/60_xfx777be_pcb_big.jpg

And there's nothing all that interesting on the secondary side (solder
side). The big chips there, could be digital format conversion,
going from proprietary GPU video out bus, to dual link DVI on a
connector. Something like that. Or maybe it's a VGA output circuit.

http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/56_xfx777be_frr_big.jpg

The thing uses Polymer caps (no stress relief lines on top).
So it's not likely to be a leaky cap problem. The odds of you
spotting a visual problem, are just about zero. If the problem
was going to be that easy to diagnose, it would probably
be smoking by now.

This is the card in question;
http://s308.photobucket.com/user/misfitnz/library/Asus HD 7770 The area
that the close-up is of gets much too hot to touch within a few seconds of
switching the computer on and listening through a thin tube the sound seems
to come from there.

I see that there's a place for a capacitor that is empty right there - is
this another case on skimping on parts to save 50c only to have a $100
assembly fail? I have a collection of as-new solid polymer Sanyo capacitors
that I scavenged off an Intel server mobo I was given*. However the only
ones of a similar or higher value won't fit as the cooler only clears the
existing caps by 2mm or so.

I'll search my donor PCB collection for a suitable cap more thoroughly
another day. For now I've re-assembled it pending advice. The rather nice
cooler will fit my existing HD 7770 so I might swap them around if I don't
manage to get the 'new' one working.

[*] It was out of a back-up server that was never put into service -
obviously they were reliable. When the time came to upgrade a friend in the
IT biz kept it, then didn't use ot for a couple years and gave it to me. By
then it was only as powerful as a low-end desktop but very very noisy so I
sold the hot-swappable PSUs as I could courier those but the server wasn't
worth what it would cost to frieght. So I 'ratted it' - unsoldered the ~50
solid caps. I've repaired two mobos using them now - replaced failed wet
'lytics with these excellent polymer caps.

Cheers,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[Paul]

Thanks for replying Paul, I appreciate someone of your expertise giving
input.

You have better be a switching power converter expert.

I'm not an expert on anything - oh, I guess I qualify as an expert on the
management of semi-closed aquatic systems. I used to get paid around $100 /
hour for designing filtration systems for public aquaria a couple of decades
ago (when that was *really* good money).

The video card
needs to make a 1 volt supply at high current, to run the core of the
GPU chip.

Close enough. 1.1v when working and 0.9v idling - maybe 80 amps or so.

Tbe easy problems, are leaking caps, because you can see them. Video
cards (the ones over $50), were the first electronics to make use of
polymer caps, so you'd hope the traditional leaky cap problem
wouldn't be present. I expect all three of my FX5200 cards, those use
regular electrolytic caps.

Yeah, back in the day I used to replace leaky cheap caps on otherwise fine
motherboards with good quality low ESR Rubycons or similar. Some of the
boards that I re-capped are still running 10+ years later. The last couple
of builds I did I made sure that the mobos had all-solid polymer capacitors
though. I just came across a bare FX5200 board in my junk box when looking
for a suitable cap and it has wet 'lytics. Surface-mount but with stress
relief stampings on the tops.

A leaking cap can fail enough, to cause an inductor (toroid, powered
iron core), to burn up. Or a MOSFET could fail. Some power converter
circuits, measure fault current, by using the channel resistance
of the MOSFET as a means to measure current flow. Some of the
high end cards, use Volterra designs, and the principles of those
are different enough, it probably doesn't match your symptoms.
(Nothing on a Volterra, would be in audible range.)

An electronics whizz who I used to know told me that almost always when an
eltrical device makes noise it's vibrations in inductor coils and that's why
the better high frequency ones are epoxy potted. That's why I suspected it
was an inductor - though *why* it's singing is the question I guess....

Even if I had a decent scope handy, I doubt I could understand
what was wrong with one. Sure, a stuck-at fault, a short to
ground, I'd get that. But not a slight parametric problem.
I wouldn't know what was "normal".

*******

If you're absolutely determined to get $7 worth of fun out of
it. remove the heatsink, find the regulator IC, get the part number
off it, and download the PDF datasheet. That'll give you something
technical to work with. If you're lucky, the regulator is on
the solder side, and might be visible without any work.

I don't mind some 'work'. I'm a very poor invalid with nothing but time.

OK, in the first picture, the core supply is at least three phase,
and uses small dual MOSFETs in 8 pin packages (cheep). And I cannot
locate a regulator chip! That's on the left hand side of the picture.
Some lower current regulators are located on the right hand side,
and those should be for memory voltages. Maybe one regulator on the
right is for VCC and one for VTT (terminator with push/pull
capability).
http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/60_xfx777be_pcb_big.jpg

And there's nothing all that interesting on the secondary side (solder
side). The big chips there, could be digital format conversion,
going from proprietary GPU video out bus, to dual link DVI on a
connector. Something like that. Or maybe it's a VGA output circuit.

http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/56_xfx777be_frr_big.jpg

The thing uses Polymer caps (no stress relief lines on top).
So it's not likely to be a leaky cap problem. The odds of you
spotting a visual problem, are just about zero. If the problem
was going to be that easy to diagnose, it would probably
be smoking by now.

This is the card in question;
http://s308.photobucket.com/user/misfitnz/library/Asus HD 7770 The area
that the close-up is of gets much too hot to touch within a few seconds of
switching the computer on and listening through a thin tube the sound seems
to come from there.

I see that there's a place for a capacitor that is empty right there - is
this another case on skimping on parts to save 50c only to have a $100
assembly fail? I have a collection of as-new solid polymer Sanyo capacitors
that I scavenged off an Intel server mobo I was given*. However the only
ones of a similar or higher value won't fit as the cooler only clears the
existing caps by 2mm or so.

I'll search my donor PCB collection for a suitable cap more thoroughly
another day. For now I've re-assembled it pending advice. The rather nice
cooler will fit my existing HD 7770 so I might swap them around if I don't
manage to get the 'new' one working.

[*] It was out of a back-up server that was never put into service -
obviously they were reliable. When the time came to upgrade a friend in the
IT biz kept it, then didn't use ot for a couple years and gave it to me. By
then it was only as powerful as a low-end desktop but very very noisy so I
sold the hot-swappable PSUs as I could courier those but the server wasn't
worth what it would cost to frieght. So I 'ratted it' - unsoldered the ~50
solid caps. I've repaired two mobos using them now - replaced failed wet
'lytics with these excellent polymer caps.

Cheers,

On another HD7770 board, I can see room for six caps. But two of the
cap footprints are "dual footprint". It looks like two of the caps
could contain smaller caps (the two small ones give the same capacitance
as one larger one). And perhaps the alternative is to use the
one larger cap, and leave the other cap site as a de-pop. It's strange
to see them tuning the capacitance like that, if that's what they're
up to. Normally, when you check out a datasheet for a switching regulator,
the absolute value of the capacitance isn't all that critical.

The HD7770 (possibly reference design) is supposed to use a UPI 1609
regulator chip. And that's the chip I can't see in any of the photos.
It should be pretty small, a quad flat pack with small legs.

You can see the chip in a picture half way down this page.

http://www.techpowerup.com/reviews/Gigabyte/HD_7770_OC/4.html

The chip manufacturer is a waste of time. A site search isn't digging
up anything. All we get is a one line set of parameters.

http://www.upi-semi.com/product/ProductList.aspx?ArchID=140

uP1609 3 Phase Buck Controller 10.8 to 13.2V DCR VQFN6x6-40

There is no datasheet or application note on that particular page.
The overcurrent protection is via "DCR", meaning a discrete resistor.
You can see in the listing, they have some other products with "RON"
detection, which uses the MOSFET channel. The drive is "embedded",
which means they don't use outboard 8 pin chips to buffer up the
gate drive for the MOSFETs. That also makes the regulator chip
run quite hot. Sometimes those buffers have to drive a 3000pF gate
at some high frequency. The 8 pin chip method (one chip per phase) was
invented to distribute the heat, so the main regulator doesn't get
quite so hot. Using "embedded" as a method, is for cost reasons.
Fewer chips, but runs hotter. To be successful doing that, the
regulator could have a "thermal slug" on the bottom, soldered
to the board.

The 1609 is mentioned here, as an adjustable regulator (for the cards
that come from the factory with an overclock). It's possible the term
"SVID" stands for Soft Voltage ID and the ability to program a small
change to the voltage.

http://www.guru3d.com/news-story/download-msi-afterburner-3-beta-7.html

Without a datasheet, I can't really say much about how it works. It's
a buck converter, it runs off nominal 12V, and using 12V should give some
nice drive signals, so the MOSFETs are saturated well (low Rdss_on).

Paul

 

[~misfit~]

Once upon a time on usenet ~misfit~ wrote:
[snip]

This is the card in question;
http://s308.photobucket.com/user/misfitnz/library/Asus HD 7770 The

Damn URL didn't work. Try this;
http://s308.photobucket.com/user/misfitnz/library/Asus HD 7770?sort=3&page=1

Cheers,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[~misfit~]

~misfit~ wrote:
I have an HD7770 graphics card in my desktop. Just a basic one but
a big step up from the HD5670 I was using before for light gaming.

Last week I saw an auction for a faulty one on TM
http://www.trademe.co.nz/Browse/Listing.aspx?id=803566559 and put
in an autobid of $6.66 and won it. I thought that, at the very least
I'd have a really good direct-contact heatpipe cooler that I might
be able to graft onto my card (which has a simple ali block HS).
Also though, being the eternal optomist (despite life proving that
it's an unrealistic outlook) I hoped that I might be able to find a
fault with it and fix it. It arrived today and I fitted it to my
spare / guest machine in
place of the HD5670 that was in there. Much to my delight it
displayed the BIOS screen just fine then booted into Win7 (albeit
at a low resolution). A Windows popup appeared saying 'installing
drivers' and about that time I noticed a sound that was akin to a
very small mouse getting a filling. It was very quiet and I might
have missed it if I'd fitted the cover back on the PC.... The
machine installed the new drivers fine and asked to reboot - I
clicked OK. It started up at full resolution! However the tiny
mouse was still getting dental work done - maybe moreso. I started Path
of Exile and it downloaded updates (I hadn't used the machine for a
while) and then the load screen came up - and the monitor went
blank. :-( However the fact that it ran for ~10 minutes makes me
hopeful. I
can't understand why the owner didn't get a replacement under
warranty - it's an ASUS after all and the card isn't *that* old.
<whirr click clang.> Oh, unless they did and I now have the faulty
one. :-/ I did a dirty shut-down of the machine then came to my
laptop to post this and ask for advice. Am I right in thinking
that it may be an inductor (or possibly capacitor) that's faulty? I can
likely
replace either if it is and I can get a suitable replacement but I
don't want to go blindly replacing components. I was unable to
pin-point the sound because of the large shroud and heatsink /
radiator covering the PCB. Frankly I'm surprised that I heard it as
I'm sure I have age-related hearing loss of higher frequencies.
What do you folks think? Any relevant advice? I've replaced mobo
caps before and even done a bit of surface-mount reworking - about
as much as you can do without a hot-air station at least. I'd be
over the moon if I could repair it - it'd likely be the only Xmas
present I get. Cheers,

Thanks for replying Paul, I appreciate someone of your expertise
giving input.

You have better be a switching power converter expert.

I'm not an expert on anything - oh, I guess I qualify as an expert
on the management of semi-closed aquatic systems. I used to get paid
around $100 / hour for designing filtration systems for public
aquaria a couple of decades ago (when that was *really* good money).

The video card
needs to make a 1 volt supply at high current, to run the core of
the GPU chip.

Close enough. 1.1v when working and 0.9v idling - maybe 80 amps or
so.

Tbe easy problems, are leaking caps, because you can see them. Video
cards (the ones over $50), were the first electronics to make use of
polymer caps, so you'd hope the traditional leaky cap problem
wouldn't be present. I expect all three of my FX5200 cards, those
use regular electrolytic caps.

Yeah, back in the day I used to replace leaky cheap caps on
otherwise fine motherboards with good quality low ESR Rubycons or
similar. Some of the boards that I re-capped are still running 10+
years later. The last couple of builds I did I made sure that the
mobos had all-solid polymer capacitors though. I just came across a
bare FX5200 board in my junk box when looking for a suitable cap and
it has wet 'lytics. Surface-mount but with stress relief stampings
on the tops.

A leaking cap can fail enough, to cause an inductor (toroid, powered
iron core), to burn up. Or a MOSFET could fail. Some power converter
circuits, measure fault current, by using the channel resistance
of the MOSFET as a means to measure current flow. Some of the
high end cards, use Volterra designs, and the principles of those
are different enough, it probably doesn't match your symptoms.
(Nothing on a Volterra, would be in audible range.)

An electronics whizz who I used to know told me that almost always
when an eltrical device makes noise it's vibrations in inductor
coils and that's why the better high frequency ones are epoxy
potted. That's why I suspected it was an inductor - though *why*
it's singing is the question I guess....

Even if I had a decent scope handy, I doubt I could understand
what was wrong with one. Sure, a stuck-at fault, a short to
ground, I'd get that. But not a slight parametric problem.
I wouldn't know what was "normal".

*******

If you're absolutely determined to get $7 worth of fun out of
it. remove the heatsink, find the regulator IC, get the part number
off it, and download the PDF datasheet. That'll give you something
technical to work with. If you're lucky, the regulator is on
the solder side, and might be visible without any work.

I don't mind some 'work'. I'm a very poor invalid with nothing but
time.

OK, in the first picture, the core supply is at least three phase,
and uses small dual MOSFETs in 8 pin packages (cheep). And I cannot
locate a regulator chip! That's on the left hand side of the
picture. Some lower current regulators are located on the right
hand side, and those should be for memory voltages. Maybe one regulator
on the
right is for VCC and one for VTT (terminator with push/pull
capability).
http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/60_xfx777be_pcb_big.jpg

And there's nothing all that interesting on the secondary side
(solder side). The big chips there, could be digital format
conversion, going from proprietary GPU video out bus, to dual link DVI
on a
connector. Something like that. Or maybe it's a VGA output circuit.

http://www.xbitlabs.com/picture/?sr...-7870-hd-7850-hd-7770/56_xfx777be_frr_big.jpg

The thing uses Polymer caps (no stress relief lines on top).
So it's not likely to be a leaky cap problem. The odds of you
spotting a visual problem, are just about zero. If the problem
was going to be that easy to diagnose, it would probably
be smoking by now.

This is the card in question;
http://s308.photobucket.com/user/misfitnz/library/Asus HD 7770 The
area that the close-up is of gets much too hot to touch within a few
seconds of switching the computer on and listening through a thin
tube the sound seems to come from there.

I see that there's a place for a capacitor that is empty right there
- is this another case on skimping on parts to save 50c only to have
a $100 assembly fail? I have a collection of as-new solid polymer
Sanyo capacitors that I scavenged off an Intel server mobo I was
given*. However the only ones of a similar or higher value won't fit
as the cooler only clears the existing caps by 2mm or so.

I'll search my donor PCB collection for a suitable cap more
thoroughly another day. For now I've re-assembled it pending advice.
The rather nice cooler will fit my existing HD 7770 so I might swap
them around if I don't manage to get the 'new' one working.

[*] It was out of a back-up server that was never put into service -
obviously they were reliable. When the time came to upgrade a friend
in the IT biz kept it, then didn't use ot for a couple years and
gave it to me. By then it was only as powerful as a low-end desktop
but very very noisy so I sold the hot-swappable PSUs as I could
courier those but the server wasn't worth what it would cost to
frieght. So I 'ratted it' - unsoldered the ~50 solid caps. I've
repaired two mobos using them now - replaced failed wet 'lytics with
these excellent polymer caps. Cheers,

On another HD7770 board, I can see room for six caps. But two of the
cap footprints are "dual footprint". It looks like two of the caps
could contain smaller caps (the two small ones give the same
capacitance as one larger one). And perhaps the alternative is to use
the one larger cap, and leave the other cap site as a de-pop. It's strange
to see them tuning the capacitance like that, if that's what they're
up to. Normally, when you check out a datasheet for a switching
regulator, the absolute value of the capacitance isn't all that
critical.
The HD7770 (possibly reference design) is supposed to use a UPI 1609
regulator chip. And that's the chip I can't see in any of the photos.
It should be pretty small, a quad flat pack with small legs.

You can see the chip in a picture half way down this page.

http://www.techpowerup.com/reviews/Gigabyte/HD_7770_OC/4.html

The chip manufacturer is a waste of time. A site search isn't digging
up anything. All we get is a one line set of parameters.

http://www.upi-semi.com/product/ProductList.aspx?ArchID=140

uP1609 3 Phase Buck Controller 10.8 to 13.2V DCR VQFN6x6-40
There is no datasheet or application note on that particular page.
The overcurrent protection is via "DCR", meaning a discrete resistor.
You can see in the listing, they have some other products with "RON"
detection, which uses the MOSFET channel. The drive is "embedded",
which means they don't use outboard 8 pin chips to buffer up the
gate drive for the MOSFETs. That also makes the regulator chip
run quite hot. Sometimes those buffers have to drive a 3000pF gate
at some high frequency. The 8 pin chip method (one chip per phase) was
invented to distribute the heat, so the main regulator doesn't get
quite so hot. Using "embedded" as a method, is for cost reasons.
Fewer chips, but runs hotter. To be successful doing that, the
regulator could have a "thermal slug" on the bottom, soldered
to the board.

The 1609 is mentioned here, as an adjustable regulator (for the cards
that come from the factory with an overclock). It's possible the term
"SVID" stands for Soft Voltage ID and the ability to program a small
change to the voltage.

http://www.guru3d.com/news-story/download-msi-afterburner-3-beta-7.html

Without a datasheet, I can't really say much about how it works. It's
a buck converter, it runs off nominal 12V, and using 12V should give
some nice drive signals, so the MOSFETs are saturated well (low
Rdss_on).

Hi Paul. Sorry about the delay replying - pain's been a bitch lately.

I've uploaded a couple more pics to my photobucket. I think I found the
regulator and nearby perhaps "DCR" blown? The blown component is very very
small and about at the limit of what I might be able to solder - assuming
the PCB is still all right and I had a clue as to what value component to
use as a replacement. (If I'm going to replace it I'll use a non-SM
component and solder the leads to the pads for the SMC, then glue it down.)

Any ideas as to what I should try to solder there if I can clean it up?
Cheers,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[~misfit~]

Once upon a time on usenet Paul wrote: [snip]

The HD7770 (possibly reference design) is supposed to use a UPI 1609
regulator chip. And that's the chip I can't see in any of the photos.
It should be pretty small, a quad flat pack with small legs.

You can see the chip in a picture half way down this page.

http://www.techpowerup.com/reviews/Gigabyte/HD_7770_OC/4.html

The chip manufacturer is a waste of time. A site search isn't digging
up anything. All we get is a one line set of parameters.

http://www.upi-semi.com/product/ProductList.aspx?ArchID=140

uP1609 3 Phase Buck Controller 10.8 to 13.2V DCR VQFN6x6-40
There is no datasheet or application note on that
particular page. The overcurrent protection is via "DCR", meaning a
discrete resistor.
You can see in the listing, they have some other products with "RON"
detection, which uses the MOSFET channel. The drive is "embedded",
which means they don't use outboard 8 pin chips to buffer up the
gate drive for the MOSFETs. That also makes the regulator chip
run quite hot. Sometimes those buffers have to drive a 3000pF gate
at some high frequency. The 8 pin chip method (one chip per phase)
was invented to distribute the heat, so the main regulator doesn't
get quite so hot. Using "embedded" as a method, is for cost reasons.
Fewer chips, but runs hotter. To be successful doing that, the
regulator could have a "thermal slug" on the bottom, soldered
to the board.

The 1609 is mentioned here, as an adjustable regulator (for the cards
that come from the factory with an overclock). It's possible the term
"SVID" stands for Soft Voltage ID and the ability to program a small
change to the voltage.

http://www.guru3d.com/news-story/download-msi-afterburner-3-beta-7.html

Without a datasheet, I can't really say much about how it works. It's
a buck converter, it runs off nominal 12V, and using 12V should give
some nice drive signals, so the MOSFETs are saturated well (low
Rdss_on).

Hi Paul. Sorry about the delay replying - pain's been a bitch lately.

I've uploaded a couple more pics to my photobucket. I think I found
the regulator and nearby perhaps "DCR" blown? The blown component is
very very small and about at the limit of what I might be able to
solder - assuming the PCB is still all right and I had a clue as to
what value component to use as a replacement. (If I'm going to
replace it I'll use a non-SM component and solder the leads to the
pads for the SMC, then glue it down.)
Any ideas as to what I should try to solder there if I can clean it
up? Cheers,

D'oh! It would help if I included links to the pics yes?
http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-a.jpg.html?sort=3&o=0
http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-b.jpg.html?sort=3&o=1
http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-c.jpg.html?sort=3&o=2

Cheers,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[~misfit~]

Once upon a time on usenet ~misfit~ wrote:
[snip]

D'oh! It would help if I included links to the pics yes?
http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-a.jpg.html?sort=3&o=0
http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-b.jpg.html?sort=3&o=1
http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-c.jpg.html?sort=3&o=2

So I got out the 99% iso and Q-Tips to clean the failed component;

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-d.jpg.html?sort=3&o=3

... and noticed that the corner of the VRM was different, as if some glue had
been removed (my camera is better than my eyes);

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-e.jpg.html?sort=3&o=4
http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-f.jpg.html?sort=3&o=2

So I got the iso and Q-Tips out again and cleaned the VRM somewhat;

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-g.jpg.html?sort=3&o=1

The writing was still not clear so I had another go with the iso;

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-g.jpg.html?sort=3&o=1

So it seems that's the VRM on this card and that it may have had a heatsink
glued to it once upon a time. There was quite a think layer of glue that
took some disolving so I think maybe it was *insulating* the chip which got
hot then popped the small SMC - maybe...

All of which lends more credence to my thinking that this card may be a
warranty replacement. The seller sells a heck of a lot of computer
components, quite a few of which are sold as being faulty and he doesn't
answer questions on auctions.

LOL, I'm staring to wonder if I can't get this going whether it might be
worth approaching ASUS NZ and telling them I bought this card and it's
failed and surely it should be under warranty? At the very least them
checking the serial number might plug the hole in their warehouse whereby
warranty returns are making their way back into the consumer space and
potentially damaging their reputation.

That said I'd rather get this thing working if possible. Do you think it's
possible Paul?

Cheers,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[~misfit~]

Once upon a time on usenet ~misfit~ wrote:
[snip]

So it seems that's the VRM on this card and that it may have had a
heatsink glued to it once upon a time. There was quite a think layer
of glue that took some disolving so I think maybe it was *insulating*
the chip which got hot then popped the small SMC - maybe...

I found a pic of the back of the card an Newegg and there isn't a heatsink
on that card;
http://www.newegg.com/Product/Product.aspx?Item=N82E16814121632

So I wonder what the crud on the VRM of my card was? Maybe someone tried to
put a heatsink on it when it played up, after they noticed that it got very
hot. Then they took it off again (or it fell off) before they sold it.

I'd love to be able to get this going - I hate waste - not to mention I'm
poverty-stricken and it would be a real win for me.

Oh, only just noticed the typo in the subject. Then as now my back pain is
getting to the stage where I can no longer sit up and so my typing
suffers....
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[Paul]

Once upon a time on usenet ~misfit~ wrote:
[snip]

So it seems that's the VRM on this card and that it may have had a
heatsink glued to it once upon a time. There was quite a think layer
of glue that took some disolving so I think maybe it was *insulating*
the chip which got hot then popped the small SMC - maybe...

I found a pic of the back of the card an Newegg and there isn't a heatsink
on that card;
http://www.newegg.com/Product/Product.aspx?Item=N82E16814121632

So I wonder what the crud on the VRM of my card was? Maybe someone tried to
put a heatsink on it when it played up, after they noticed that it got very
hot. Then they took it off again (or it fell off) before they sold it.

I'd love to be able to get this going - I hate waste - not to mention I'm
poverty-stricken and it would be a real win for me.

Oh, only just noticed the typo in the subject. Then as now my back pain is
getting to the stage where I can no longer sit up and so my typing
suffers....

The blue-green residue looks like copper chloride to me.
That looks like a chemical reaction. I've never seen
a surface mount component that looked quite like that before.

As for the chip, my theory was that your board would
use the same regulator as the more-or-less reference
design. Instead, the Asus design uses what I would
consider to be an Asus chip. I don't know who actually
makes the Digi+ VRM, and I wasn't able to unearth any
details in a search. When used on motherboards, that
chip has supposedly both thermal and current sense
capability. But now we don't know if it does RON (RDSS_ON
channel resistance) or DCR (current sense by low ohms
sense resistor).

It's possible if a MOSFET failed in a particular
state, if a sense resistor was present, it would
get blown to hell. And I don't think that's what
happened to that resistor. I think someone may have
dropped a dab of something corrosive on it. But the
resistor body should not have degraded. I don't know
about you, but I find epoxy resists a lot of the things
I've poured on it. And it's even used as a material to
make pH probes, implying it can take both acids and
bases. Now, maybe it's burned, but the body of a
surface mount component has to take a pretty high
temperature during the solder profile. And the other
components in the area don't particularly look heat damaged.

And how exactly does a digi+ vrm work. I can't find
a damn thing that explains the theory of operation.
So that's going to make it pretty hard to guess
what that particular resistor is doing.

The 41C marking on the corroded part, could be an EIA96 code.
It looks like 261*100 = 26.1K ohms. I gather it's a resistor,
because it doesn't "look ceramic" like the light brown one
next to it. The 26.1K could be a feedback resistor,
shaping gain or frequency response.

http://www.hobby-hour.com/electronics/smdcalc.php

If I was working on that, I'd

1) Unsolder the 41C.
2) Check with ohmmeter, and see if it is 26.1K.
If so, chances are it's not the faulty bit.
3) Clean the site thoroughly. Wipe clean with isopropyl
and hope it gets the water soluble inorganic material
causing the corrosion. Check to see if the copper
pads on either end of the resistor have been
completely eaten away. If so, the regulator may
have been commanded to go nuts, and burn up the GPU.
4) Solder a new 41C in place. I can't tell from here
if that is an 0603 or an 0805 (body size).
5) Continue looking for damage.
6) Don't power up again, until exhausting all avenues.

Paul

 

[~misfit~]

Once upon a time on usenet ~misfit~ wrote:
[snip]

So it seems that's the VRM on this card and that it may have had a
heatsink glued to it once upon a time. There was quite a think layer
of glue that took some disolving so I think maybe it was
*insulating* the chip which got hot then popped the small SMC -
maybe...

I found a pic of the back of the card an Newegg and there isn't a
heatsink on that card;
http://www.newegg.com/Product/Product.aspx?Item=N82E16814121632

So I wonder what the crud on the VRM of my card was? Maybe someone
tried to put a heatsink on it when it played up, after they noticed
that it got very hot. Then they took it off again (or it fell off)
before they sold it. I'd love to be able to get this going - I hate
waste - not to
mention I'm poverty-stricken and it would be a real win for me.

Oh, only just noticed the typo in the subject. Then as now my back
pain is getting to the stage where I can no longer sit up and so my
typing suffers....

The blue-green residue looks like copper chloride to me.
That looks like a chemical reaction. I've never seen
a surface mount component that looked quite like that before.

I thought it odd too - almost as if something above it in the machine it was
originally fitted to leaked - maybe an electrolytic cap? That could have
been what was on the VRM. All the tiny fibres in the verdigris gunk could be
from someone trying to clean the card up. All gueswork though.

As for the chip, my theory was that your board would
use the same regulator as the more-or-less reference
design. Instead, the Asus design uses what I would
consider to be an Asus chip. I don't know who actually
makes the Digi+ VRM, and I wasn't able to unearth any
details in a search. When used on motherboards, that
chip has supposedly both thermal and current sense
capability. But now we don't know if it does RON (RDSS_ON
channel resistance) or DCR (current sense by low ohms
sense resistor).

Asus sell this card as 'tweakable' - it's designed to overclock, they
include a software tweaking tool on the CD with the card (which I don't
have). That's likely why they have the non-standard VRM?

It's possible if a MOSFET failed in a particular
state, if a sense resistor was present, it would
get blown to hell. And I don't think that's what
happened to that resistor. I think someone may have
dropped a dab of something corrosive on it. But the
resistor body should not have degraded. I don't know
about you, but I find epoxy resists a lot of the things
I've poured on it. And it's even used as a material to
make pH probes, implying it can take both acids and
bases. Now, maybe it's burned, but the body of a
surface mount component has to take a pretty high
temperature during the solder profile. And the other
components in the area don't particularly look heat damaged.

No, they don't - and that area of the card gets *very* hot before it shuts
down - well over 150ºC according to my finger.

And how exactly does a digi+ vrm work. I can't find
a damn thing that explains the theory of operation.
So that's going to make it pretty hard to guess
what that particular resistor is doing.

The 41C marking on the corroded part, could be an EIA96 code.
It looks like 261*100 = 26.1K ohms. I gather it's a resistor,
because it doesn't "look ceramic" like the light brown one
next to it. The 26.1K could be a feedback resistor,
shaping gain or frequency response.

http://www.hobby-hour.com/electronics/smdcalc.php

If I was working on that, I'd

1) Unsolder the 41C.
2) Check with ohmmeter, and see if it is 26.1K.
If so, chances are it's not the faulty bit.
3) Clean the site thoroughly. Wipe clean with isopropyl
and hope it gets the water soluble inorganic material
causing the corrosion. Check to see if the copper
pads on either end of the resistor have been
completely eaten away. If so, the regulator may
have been commanded to go nuts, and burn up the GPU.
4) Solder a new 41C in place. I can't tell from here
if that is an 0603 or an 0805 (body size).
5) Continue looking for damage.
6) Don't power up again, until exhausting all avenues.

Thanks heaps for the input Paul - I very much appreciate it. I just got the
micrometer out and the component in question is 2mm x 1mm. That's smaller
than my previous smallest previous SM replacement work (a 3mm x 1.5mm fuse
on a ThinkPad planar, a common failure on T43s if the power adapater brick
cord gets damaged). I'm not sure that the finest tip I have for my soldering
iron will do the job (especially being lead-free solder it's hard to get
enough heat to the pad with a fine tip and I don't have a hot air rework
station).

That said I'll try - I have little to lose at this stage. I'll 'soak' the
area in iso before I start and remove as much gunk as I can with a
toothpick. Fingers crossed. When I have some low-pain time to devote to it
I'll attempt it. Thanks again for your input - I likely wouldn't even try
without it.

Best regards,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[Paul]

Once upon a time on usenet ~misfit~ wrote:
[snip]
So it seems that's the VRM on this card and that it may have had a
heatsink glued to it once upon a time. There was quite a think layer
of glue that took some disolving so I think maybe it was
*insulating* the chip which got hot then popped the small SMC -
maybe...
I found a pic of the back of the card an Newegg and there isn't a
heatsink on that card;
http://www.newegg.com/Product/Product.aspx?Item=N82E16814121632

So I wonder what the crud on the VRM of my card was? Maybe someone
tried to put a heatsink on it when it played up, after they noticed
that it got very hot. Then they took it off again (or it fell off)
before they sold it. I'd love to be able to get this going - I hate
waste - not to
mention I'm poverty-stricken and it would be a real win for me.

Oh, only just noticed the typo in the subject. Then as now my back
pain is getting to the stage where I can no longer sit up and so my
typing suffers....

The blue-green residue looks like copper chloride to me.
That looks like a chemical reaction. I've never seen
a surface mount component that looked quite like that before.

I thought it odd too - almost as if something above it in the machine it was
originally fitted to leaked - maybe an electrolytic cap? That could have
been what was on the VRM. All the tiny fibres in the verdigris gunk could be
from someone trying to clean the card up. All gueswork though.

As for the chip, my theory was that your board would
use the same regulator as the more-or-less reference
design. Instead, the Asus design uses what I would
consider to be an Asus chip. I don't know who actually
makes the Digi+ VRM, and I wasn't able to unearth any
details in a search. When used on motherboards, that
chip has supposedly both thermal and current sense
capability. But now we don't know if it does RON (RDSS_ON
channel resistance) or DCR (current sense by low ohms
sense resistor).

Asus sell this card as 'tweakable' - it's designed to overclock, they
include a software tweaking tool on the CD with the card (which I don't
have). That's likely why they have the non-standard VRM?

It's possible if a MOSFET failed in a particular
state, if a sense resistor was present, it would
get blown to hell. And I don't think that's what
happened to that resistor. I think someone may have
dropped a dab of something corrosive on it. But the
resistor body should not have degraded. I don't know
about you, but I find epoxy resists a lot of the things
I've poured on it. And it's even used as a material to
make pH probes, implying it can take both acids and
bases. Now, maybe it's burned, but the body of a
surface mount component has to take a pretty high
temperature during the solder profile. And the other
components in the area don't particularly look heat damaged.

No, they don't - and that area of the card gets *very* hot before it shuts
down - well over 150ºC according to my finger.

And how exactly does a digi+ vrm work. I can't find
a damn thing that explains the theory of operation.
So that's going to make it pretty hard to guess
what that particular resistor is doing.

The 41C marking on the corroded part, could be an EIA96 code.
It looks like 261*100 = 26.1K ohms. I gather it's a resistor,
because it doesn't "look ceramic" like the light brown one
next to it. The 26.1K could be a feedback resistor,
shaping gain or frequency response.

http://www.hobby-hour.com/electronics/smdcalc.php

If I was working on that, I'd

1) Unsolder the 41C.
2) Check with ohmmeter, and see if it is 26.1K.
If so, chances are it's not the faulty bit.
3) Clean the site thoroughly. Wipe clean with isopropyl
and hope it gets the water soluble inorganic material
causing the corrosion. Check to see if the copper
pads on either end of the resistor have been
completely eaten away. If so, the regulator may
have been commanded to go nuts, and burn up the GPU.
4) Solder a new 41C in place. I can't tell from here
if that is an 0603 or an 0805 (body size).
5) Continue looking for damage.
6) Don't power up again, until exhausting all avenues.

Thanks heaps for the input Paul - I very much appreciate it. I just got the
micrometer out and the component in question is 2mm x 1mm. That's smaller
than my previous smallest previous SM replacement work (a 3mm x 1.5mm fuse
on a ThinkPad planar, a common failure on T43s if the power adapater brick
cord gets damaged). I'm not sure that the finest tip I have for my soldering
iron will do the job (especially being lead-free solder it's hard to get
enough heat to the pad with a fine tip and I don't have a hot air rework
station).

That said I'll try - I have little to lose at this stage. I'll 'soak' the
area in iso before I start and remove as much gunk as I can with a
toothpick. Fingers crossed. When I have some low-pain time to devote to it
I'll attempt it. Thanks again for your input - I likely wouldn't even try
without it.

Best regards,

Some more breadcrumb material for you...

I decided to try a couple more Digi+ VRM searches.
Now before, I got nothing but marketing blurbs, which was discouraging.

And then I saw this. Still not a breakthru at all.
"DIGI+ ASP1108", "DIGI+ VRM RPU ASP1107", "ASUS TPU"

http://www.modders-inc.com/asus-a88x-pro-motherboard-review/2/

The ASP tends to mean "Application Specific Processor", a flavor
of customization.

The legends on the chip still weren't giving anything away.
More artsy than proprietary.

So I popped that kind of stuff back into the search engine,
and tried again. And I found this in Russian. It had DIGI and
ASP in it. And they gave a datasheet link.

http://www.rom.by/forum/Ishchu_Shim_CPU_Digi_VRM_ASP1000C-A62-Ukraina

Now, this site is International Rectifier. Maker of MOSFETs.
And what better candidate than to own some VRM chippage.
The branding on the actual datasheet, is "CHiL", implying
IRF bought them or something.

http://irf.com/product-info/datasheets/data/pb-chl8326.pdf

That one has temperature sensing (perhaps for overall overheating)
as well as current sense. The current sense looks like differential
sensing across the inductor coil. Copper has a significant
temperature coefficient of resistance, which is why measuring
V=R*I across Rcoil is problematic. But, if you include temperature
sensing, you can compensate for the coil temperature.

Now, since Asus has at least three of those chips, with custom
branding printed on top of them, I don't know which chip that
is in the family. But at least it gives you a possible breadcrumb
as to where that chip came from.

You could start by doing a pin count on your Digi, and see
if it matches the pins on that one.

When they do current sense across a coil, they use a sample
and hold circuit, and typically wait to a certain time point
in the switching regulator operation, before taking a snapshot
of the Vcoil level. And that measure is then supposed to be
proportional to how hard the phase is working, or whether the
phase is "overcurrent" or not. On overcurrent, the gate drive
needs to be turned off on all MOSFETs.

That particular chip, on page 2, appears non-embedded.
External CHL8510 chips drive the MOSFET gates, to isolate
gate loading from the main chip. (So the main chip won't
overheat.) The PWM signal from the main chip, when passed to
the buffering chip, also creates two gate control signals
to turn on the correct MOSFET in the totem pole.

*******

This press release announces the acquisition of CHiL by IRF
(a.k.a International Rectifier).

http://www.businesswire.com/news/ho...-Acquires-CHiL-Semiconductor-Leading-Supplier

Paul

 

[~misfit~]

Once upon a time on usenet Paul wrote:

[snip saved info]

Thank you very much for that Paul, I really appreciate your interest. The
chip on my graphics card has nine legs per side whereas those in that pdf
you found are 12 and 14 respectively. :-/

I logged in to my RS account last week and ordered 50 of each (minimum
order) 0.1w 26.1K 0603 SMD and 0.25w through-mount 26.1K resistors. I got a
notification email the next day (Wednesday) saying they'd been dispatched.
In the past that means I get them usually the next day and at the latest two
days later. (I live on the southern outskirts of NZ's biggest city,
Auckland.)

When I hadn't received them by the Friday evening I checked the email and it
said they'd been sent with Geodis Wilson UK Ltd. I'd never heard of them and
knowing RS are global feared the worst. When I checked I saw that they only
have two offices in NZ, in the two largest international airports (Auckland
and Wellington). So I assume that the components are coming from outside NZ.

So I decided to get the card off my bench for now. Before I did I thought I
might as well clean that SM resistor up some more, maybe try to unsolder it.
I dipped a Q Tip in 98% isopropanol, dabbed it while it was quite wet and
left it for ~20seconds. Then I dipped it back into iso and strated to gently
wipe around the resistor area. After the first wipe it looked different and
I saw the resistor on the Q tip! My eyes aren't what they used to be (and
even if they were I'd be struggling to see the level of detail that I needed
to) so I got my trusty camera out, set it to macro (minimum focal length 5mm
and 10 megapixels) and took a few pics to check out the pads.

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-i.jpg.html

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-j.jpg.html

Not looking good, no pads left to speak of. However I notice looking at the
second pic that the traces for the missing resistor go from a via very close
to one end then to the last leg on that side of the VRM. So I thought that
maybe it's good that I ordered the through-mount resistors as well - I can
maybe clean the coating off the top of the via, cut and shape the legs of a
resistor to fit, tin them, dip them in solder paste and solder them in
place.....

Then, zooming into the pic further I noticed that the trace going from the
VRM to where the resistor was is lifted slightly. Also the traces going to
two other resistors nearby look the same... So I'm starting to think there's
less chance that the resistor failure is the primary problem (it's still not
ruled out completely but...). The board did get *extremely* hot in that
area.

What do you think looking at that second pic Paul? What are the odds that
replacing the 26.1K resistor (when I get the package from RS) will bring the
board back to life? I'm starting to think that I purchased a really good GPU
cooler rather than a possibly repairable graphics card.

Cheers - and again, thanks for your input,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[Paul]

Once upon a time on usenet Paul wrote:

[snip saved info]

Thank you very much for that Paul, I really appreciate your interest. The
chip on my graphics card has nine legs per side whereas those in that pdf
you found are 12 and 14 respectively. :-/

I logged in to my RS account last week and ordered 50 of each (minimum
order) 0.1w 26.1K 0603 SMD and 0.25w through-mount 26.1K resistors. I got a
notification email the next day (Wednesday) saying they'd been dispatched.
In the past that means I get them usually the next day and at the latest two
days later. (I live on the southern outskirts of NZ's biggest city,
Auckland.)

When I hadn't received them by the Friday evening I checked the email and it
said they'd been sent with Geodis Wilson UK Ltd. I'd never heard of them and
knowing RS are global feared the worst. When I checked I saw that they only
have two offices in NZ, in the two largest international airports (Auckland
and Wellington). So I assume that the components are coming from outside NZ.

So I decided to get the card off my bench for now. Before I did I thought I
might as well clean that SM resistor up some more, maybe try to unsolder it.
I dipped a Q Tip in 98% isopropanol, dabbed it while it was quite wet and
left it for ~20seconds. Then I dipped it back into iso and strated to gently
wipe around the resistor area. After the first wipe it looked different and
I saw the resistor on the Q tip! My eyes aren't what they used to be (and
even if they were I'd be struggling to see the level of detail that I needed
to) so I got my trusty camera out, set it to macro (minimum focal length 5mm
and 10 megapixels) and took a few pics to check out the pads.

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-i.jpg.html

http://s308.photobucket.com/user/misfitnz/media/Asus HD 7770/HD7770-j.jpg.html

Not looking good, no pads left to speak of. However I notice looking at the
second pic that the traces for the missing resistor go from a via very close
to one end then to the last leg on that side of the VRM. So I thought that
maybe it's good that I ordered the through-mount resistors as well - I can
maybe clean the coating off the top of the via, cut and shape the legs of a
resistor to fit, tin them, dip them in solder paste and solder them in
place.....

Then, zooming into the pic further I noticed that the trace going from the
VRM to where the resistor was is lifted slightly. Also the traces going to
two other resistors nearby look the same... So I'm starting to think there's
less chance that the resistor failure is the primary problem (it's still not
ruled out completely but...). The board did get *extremely* hot in that
area.

What do you think looking at that second pic Paul? What are the odds that
replacing the 26.1K resistor (when I get the package from RS) will bring the
board back to life? I'm starting to think that I purchased a really good GPU
cooler rather than a possibly repairable graphics card.

Cheers - and again, thanks for your input,

"I'm starting to think that I purchased a really good GPU
cooler rather than a possibly repairable graphics card."

Possibly.

The thing about high power circuits, is there is the initial failure,
which might be quite minor (feedback resistor disconnected). But there
could be collateral damage, such as other high power components forced
to do something they weren't supposed to do.

I can see that blue green color on a few other locations
in your picture, so damage to the copper could
exist in more than one place.

At least when you get a resistor to stuff in there, you
can give it one try and see what you get.

What I wouldn't want to happen, is the video card causing
damage to the computer, and putting you further out of pocket.

Paul

 

[Skybuck Flying]

I had a motherboard was which also "sang" to me.

It was a copper coil that was rubbing against it's plastic pole... slightly
expanding/contracting during intense memory seeking of the ram chips.

That's where I would start in your case... look for big copper coils... or
maybe it's the heatsink/copper pipes doing the same thing as heat goes
through it or other vibrations.

Or maybe it's the fan... another component which can make noises

Bye,
Skybuck.

 

[~misfit~]

cooler rather than a possibly repairable graphics card.

Possibly.

The thing about high power circuits, is there is the initial failure,
which might be quite minor (feedback resistor disconnected). But there
could be collateral damage, such as other high power components forced
to do something they weren't supposed to do.

I can see that blue green color on a few other locations
in your picture, so damage to the copper could
exist in more than one place.

At least when you get a resistor to stuff in there, you
can give it one try and see what you get.

What I wouldn't want to happen, is the video card causing
damage to the computer, and putting you further out of pocket.

I want to thank you again for your input on my issue Paul and update you...

With the damage to the pads and traces of the trashed SMD I attempted to use
a through-hole type resistor and solder it to the via and VRM leg. However
lacking a hot air re-work station and a soldering microscope (or something
that would do the job) it was an impossible task for me. The diameter of the
leads on the resistor was such that it was thicker than the gap between VRM
legs - it covers two legs and the gap between. Even cutting it at an angle I
was had major difficulty soldering it.

However the other end I had zero success with. I simply couldn't get the leg
to stick to the exposed trace / via. I couldn't get enough heat into it with
just an iron. As much as I'd like to have a rework station and suitable USB
microscope (more like this project;
http://www.weescribble.com/soldering-small-stuff-with-help-of-usb-microscope/
rather than this;
http://www.diyinhk.com/shop/usb-microscope-1080p-for-smt-soldering/48-1080p-smt-microscope.html )
.. I can't justify the credit card debt. It would cost me far more than the
card is worth and, even though I would get further use out of both a re-work
station and micrscope they are beyond me financially.

I've put the card away and one day I may be able to attempt to solder in a
replacement resitor with better equipment. However for now it remains a pipe
dream.

Happy holidays,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)

 

[Paul]

I want to thank you again for your input on my issue Paul and update you...

With the damage to the pads and traces of the trashed SMD I attempted to use
a through-hole type resistor and solder it to the via and VRM leg. However
lacking a hot air re-work station and a soldering microscope (or something
that would do the job) it was an impossible task for me. The diameter of the
leads on the resistor was such that it was thicker than the gap between VRM
legs - it covers two legs and the gap between. Even cutting it at an angle I
was had major difficulty soldering it.

However the other end I had zero success with. I simply couldn't get the leg
to stick to the exposed trace / via. I couldn't get enough heat into it with
just an iron. As much as I'd like to have a rework station and suitable USB
microscope (more like this project;
http://www.weescribble.com/soldering-small-stuff-with-help-of-usb-microscope/
rather than this;
http://www.diyinhk.com/shop/usb-microscope-1080p-for-smt-soldering/48-1080p-smt-microscope.html )
. I can't justify the credit card debt. It would cost me far more than the
card is worth and, even though I would get further use out of both a re-work
station and micrscope they are beyond me financially.

I've put the card away and one day I may be able to attempt to solder in a
replacement resitor with better equipment. However for now it remains a pipe
dream.

Happy holidays,

Well, you know how picky solder is about what it sticks to.
That green stuff wouldn't be helping matters. So even if you
had the finest equipment, the project (resuscitating corrosion),
isn't exactly easy.

I used to build PCBs at home. And one of those, my digital clock
project with vacuum fluorescent display tubes, the copper got
attacked and started making copper chloride, until it ate
right through the tracks. It took about ten years, but
the clock was ruined. That's a slightly different color than
your problem. This is what the project was similar to. Mine
had six digits too.

http://www.tubeclockdb.com/images/stories/bl-100-48-syl-tu5.jpg

Those look best, if you have a bluish plexiglass cover to put
over them. It helps hide the infrastructure and you just
see the numbers.

Paul

 

[~misfit~]

Well, you know how picky solder is about what it sticks to.
That green stuff wouldn't be helping matters. So even if you
had the finest equipment, the project (resuscitating corrosion),
isn't exactly easy.

Yeah. I cleaned in extremely well and even exposed previously-covered trace
to try to get it to stick but no luck. It probably doesn't help that my
solder paste is getting a bit old.

I used to build PCBs at home. And one of those, my digital clock
project with vacuum fluorescent display tubes, the copper got
attacked and started making copper chloride, until it ate
right through the tracks.

I can understand it if it's coastal conditions where salty air can get to
the copper. However the copper salts on my card were growing all around the
failed component and nowhere else. (The corrosion you noticed on the PCB
nearby was where I'd cleaned the area with isopropanol and it had evaporated
leaving a few tiny bits of crystaline stuff where I'd wiped it to.)

It took about ten years, but
the clock was ruined. That's a slightly different color than
your problem. This is what the project was similar to. Mine
had six digits too.

http://www.tubeclockdb.com/images/stories/bl-100-48-syl-tu5.jpg

Those look best, if you have a bluish plexiglass cover to put
over them. It helps hide the infrastructure and you just
see the numbers.

I've seen that project before on the internet. Looks cool.

Happy holidays,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)