00:00:00
hey guys buildzoid here and today we're
00:00:02
going to be doing more low effort
00:00:03
content um now I didn't think I'd be
00:00:05
making a video about this when I first
00:00:07
saw it on Reddit because I took a look
00:00:08
at this photo and I was like hm that
00:00:10
looks like a dead power stage and the
00:00:12
thing is this is an RTX 590 and RTX 59s
00:00:15
are high-end Nvidia gpus and the thing
00:00:18
about high-end Nvidia gpus is that they
00:00:20
tend to have the power stages fail like
00:00:23
a lot remember all those RTX 309s that
00:00:25
were dying trying to play new world yeah
00:00:28
they were just blowing the power stage
00:00:31
um so I am not surprised to see a photo
00:00:34
of a RTX 590 with a blown power stage
00:00:38
cuz that just seems like something a
00:00:40
5090 would eventually do at some point
00:00:43
though I will say I am surprised to see
00:00:44
it on a stxs or I mean an astral and I
00:00:48
am surprised that it happened so soon um
00:00:51
but ultimately without the card in hand
00:00:53
or more such failures it's impossible to
00:00:57
say if this was anything more than like
00:00:59
a freak man ufacturing defect
00:01:02
so yeah I didn't think I'd be making a
00:01:05
video about this so then why am I making
00:01:06
a video about it well I saw this article
00:01:09
on video cards accusing an innocent
00:01:12
capacitor of what I am 99% certain is a
00:01:15
power stage
00:01:16
explosion and I'm guessing video cards
00:01:19
got this uh capacitor accusation from
00:01:22
this comment over here this looks to be
00:01:24
a blown capacitor and this is just a
00:01:26
thing that annoys me to no end um for
00:01:29
some reason people who don't repair gpus
00:01:32
or motherboards blame capacitors for
00:01:36
like things that capacitors just don't
00:01:39
do okay like I'm assuming the reason why
00:01:43
the capacitor is being accused of what
00:01:44
is evidently a power stage explosion is
00:01:46
because of this discoloration over here
00:01:49
but I am relatively certain that that
00:01:51
discoloration is actually the internals
00:01:53
vaporized internals of a power stage I
00:01:55
guess vaporized and recondensed
00:01:57
internals of a power stage um so um yeah
00:02:02
that that capacitor is not at fault here
00:02:05
the thing is uh in my exper like I've
00:02:07
repaired a few gpus I've repaired so
00:02:10
many gpus I can't actually tell you how
00:02:12
many I've repaired because I've lost
00:02:13
track of them I've repaired a couple
00:02:15
motherboards I have never repaired a GPU
00:02:19
or a motherboard by replacing a solid
00:02:23
polymer capacitor because these are
00:02:25
basically indestructible at least if you
00:02:27
use them
00:02:28
properly there's like three ways you can
00:02:31
get these to die um and they're
00:02:34
basically like so one you could install
00:02:37
them backwards and that is a thing that
00:02:38
Asus did on a couple motherboards they
00:02:40
had a a batch of motherboards where this
00:02:43
style of polymer capacitor got installed
00:02:45
backwards and when you do that what
00:02:47
happens is these capacitors instead of
00:02:49
blocking DC current they just conduct DC
00:02:52
current so they pull a like you
00:02:54
basically end up with a lot of current
00:02:55
flowing through the capacitor the
00:02:57
capacitor gets extremely hot and in the
00:03:00
case of those Asus motherboards there
00:03:01
was like a mosfet in series with the
00:03:03
capacitor and that mosfet couldn't take
00:03:05
the current so that would usually burn
00:03:06
up um but ultimately that was caused by
00:03:09
the backwards capacitor just pulling
00:03:11
excessive amounts of current through a
00:03:13
mosfet that wasn't designed for it
00:03:15
finally enough the capacitor itself
00:03:17
didn't usually fail as far as I know and
00:03:20
I've actually kind of done the same
00:03:21
thing by accident myself I once
00:03:23
installed one of these like this style
00:03:25
of capacitor backwards on the Vore rail
00:03:28
of a Vega
00:03:29
um I did not know that that capacitor
00:03:31
was backwards until I pulled tried to
00:03:34
pull the card out of the system and
00:03:36
burnt my hand on the backwards capacitor
00:03:38
cuz it got just stupid hot um so um yeah
00:03:43
but the card was fine like it did not
00:03:45
care about the backwards capacitor at
00:03:47
all um but anyway um yeah so if you
00:03:52
install them backwards uh that that'll
00:03:55
you know cause them to pull excessive
00:03:56
current and that could eventually cause
00:03:58
them to blow up maybe but this one's not
00:04:00
backwards because we can just about make
00:04:02
out the polarity stripe so unless the
00:04:05
factory making these capacitors screwed
00:04:07
up the polarity stripe which if they did
00:04:10
like the card would not work you
00:04:12
wouldn't have one capacitor that's
00:04:13
backwards they'd be all backwards um and
00:04:17
other than that we also have the uh smt
00:04:19
adapter plate over here right to make
00:04:22
these capacitors uh surface mount and
00:04:26
that has the corners cut off to indicate
00:04:28
the polarity on one side and and yeah we
00:04:30
can see that corner is also cut off so
00:04:32
this thing is definitely not backwards
00:04:34
now alternatively you can also massively
00:04:36
over Vault these I once did that by
00:04:38
installing a 2 and2 volt capacitor on a
00:04:40
12vt rail I didn't read the label I
00:04:43
assumed it was 16 volts it wasn't um and
00:04:46
that produced a popping noise and that
00:04:48
was that it really didn't actually
00:04:50
produce a lot of drama um but uh yeah
00:04:53
also that 2 and2 volt capacitor was
00:04:55
backwards so I like nailed it it was
00:04:57
backwards and like 2 and 1 12 volts on
00:04:59
on 12 so um capacitor never stood a
00:05:04
chance um but uh this is a 16vt
00:05:08
capacitor I mean I don't know for sure
00:05:11
that it's 16 volts I can't actually see
00:05:13
the label but I'm 99% certain that it's
00:05:16
a 16vt capacitor because this card
00:05:18
literally doesn't have any other flavor
00:05:20
of 16vt capacitor also if it was like a
00:05:23
6.3 volts right like these are all 16
00:05:26
volts um so I don't like I don't see how
00:05:29
a like lower voltage capacitor would end
00:05:32
up on here um and if it was 6.3 volts or
00:05:37
2 and half volts the card wouldn't turn
00:05:39
on right this person was able to use the
00:05:42
card cuz the their PC randomly shut down
00:05:45
while browsing websites so evidently
00:05:48
those capacit like that capacitor can't
00:05:50
be massively underrated because if it
00:05:52
was they would have never been able to
00:05:54
turn the system on um so alternatively
00:05:57
you'd have to apply like 30 volts to it
00:05:59
which uh I don't know where you're going
00:06:00
to get that so nah and finally uh you
00:06:04
can overheat these though my experience
00:06:06
with doing that is usually using a heat
00:06:08
gun cuz you know I buy a lot of dead
00:06:10
gpus and on those dead gpus when you try
00:06:12
to heat gun off the dead power stage
00:06:14
well those dead power stages are often
00:06:16
very close to some polymer capacitors
00:06:18
and if you're not careful you can
00:06:20
accidentally cook the capacitor with the
00:06:22
hot air while trying to desolder the
00:06:24
dead power
00:06:26
stage so that's my experience with
00:06:29
overheating these and when you do that
00:06:31
what ends up what happens is if you
00:06:33
overheat them a little the casing swells
00:06:35
up cuz the thing is these are solid
00:06:36
polymers so they don't have any liquid
00:06:38
in them they're not designed to vent uh
00:06:41
gases or anything um and so if you
00:06:44
really overheat these they first puff up
00:06:46
cuz the casing doesn't have any way to
00:06:48
vent the internals if you start cooking
00:06:50
them um and if you keep cooking them
00:06:54
eventually the bottom of the capacitor
00:06:56
blows out um and I've had and like
00:07:00
sometimes that'll just lift the
00:07:01
capacitor up off the board a bit um but
00:07:04
I've also seen them literally shoot off
00:07:06
the board um so um yeah uh which is why
00:07:10
I now wear like safety glasses when
00:07:13
working with hot air near capacitors
00:07:15
because I really don't feel like getting
00:07:17
200 degree actually might be more than
00:07:19
200 degre by the time it blows off but
00:07:21
anyway I don't feel like having you know
00:07:23
hot capacitor shoot into my eye um that
00:07:27
doesn't sound very fun um
00:07:30
but anyway so you can cook these but you
00:07:32
like you really need to cook them basic
00:07:34
almost on purpose um also this doesn't
00:07:37
look swelled up and it doesn't look like
00:07:39
it lifted off the board at all so I'd
00:07:43
Hazard as a guess that this capacitor is
00:07:45
completely fine and so are all of the
00:07:47
other capacitors on this card now I have
00:07:49
repaired some motherboard actually I
00:07:51
don't think I ever repaired a
00:07:52
motherboard with a multi layer ceramic
00:07:54
but I have repaired some gpus with uh
00:07:57
cracked multi-layer ceramic capacitors
00:08:00
um so the problem with these is that
00:08:02
these are very rigid and pcbs are kind
00:08:05
of Bendy by comparison and so what
00:08:08
happens if you have like a multi-layer
00:08:09
ceramic near say a screw hole is that
00:08:12
the PCB will bend as the screw gets
00:08:15
insert like you know tighten down and
00:08:18
the capacitor won't so instead it'll
00:08:21
crack and when it cracks it internally
00:08:23
short circuits and then the next time
00:08:25
you apply power to the capacitor it
00:08:27
lights up like a little light bulb on
00:08:29
till and then well okay it depends on
00:08:32
what power supply is hooked up to it but
00:08:35
my experience is usually they end up
00:08:37
looking like a little orange LED um if
00:08:40
you have them hooked up to something
00:08:41
more substantial they might like fizzle
00:08:44
out pretty quickly um but they don't
00:08:47
tend to cause like the catastrophic kind
00:08:50
of damage that you get from Power stages
00:08:52
dying
00:08:54
um now the thing is
00:08:56
though um these are really small
00:09:00
relatively speaking like this is like
00:09:02
usually the kind of capacitor that is
00:09:04
susceptible to cracking due to PCB B
00:09:06
like the PCB getting bent is the larger
00:09:09
types like what we have over here
00:09:12
um these are really small alth so
00:09:15
there's no screw holes anywhere near
00:09:17
this now technically we do have a
00:09:19
thermal pad pushing down on the board so
00:09:22
that you know if that thermal pad is too
00:09:24
thick or too hard or both it could bend
00:09:28
the PCB
00:09:29
and then like maybe crack this cap like
00:09:31
these two over here or like these two
00:09:33
but like I doubt it that seems
00:09:36
incredibly unlikely like there are very
00:09:39
like there are cards out there with
00:09:40
extremely bent pcbs and and you don't
00:09:43
like you don't like I have like I I
00:09:46
think I repaired like two or three gpus
00:09:49
with cracked multiair ceramic capacitors
00:09:52
like this is very rare in my experience
00:09:56
so yeah also none of them had like
00:09:59
spectacular amounts of damage on there
00:10:02
um
00:10:03
so yeah I am I there's like to me this
00:10:07
is like 99% dead power stage
00:10:12
um
00:10:14
so that's my judgment and that's why I'm
00:10:16
making this video cuz it's just like it
00:10:18
just annoys me that capacitors which
00:10:21
generally don't fail and are borderline
00:10:24
indestructible unless misused get blamed
00:10:28
for power power stage failures which are
00:10:31
like pretty typical like that's the
00:10:33
thing that bugs me about this most is
00:10:35
like based on how like people just
00:10:37
always go like oh it's probably a
00:10:39
capacitor problem you'd think that like
00:10:41
capacitor failures are like common or
00:10:43
something but they're not like maybe on
00:10:45
cheap electronics that use Liquid
00:10:47
electrolyte capacitors capacitor
00:10:49
failures are a common issue but on gpus
00:10:53
and motherboards all of the capacitors
00:10:55
are some kind of solid polymer they're
00:10:57
basically indestructible
00:10:59
they don't
00:11:01
die other than the multi-layer Ceramics
00:11:03
but the multi-layer Ceramics are also
00:11:05
solid and also don't die unless you like
00:11:08
bend the board so I don't get how this
00:11:12
like how people just like always jump to
00:11:14
blaming the capacitors and it annoys me
00:11:16
because it's like it's not the freaking
00:11:18
capacitors it's like never the
00:11:21
capacitors so anyway
00:11:24
um yeah um that's the main reason I'm
00:11:27
making this video now on to why this
00:11:29
might have failed now um I don't think
00:11:32
this is a case of like Asus cheaping out
00:11:34
on the components cuz they're just using
00:11:36
Monolithic Power Systems mp8 6670 uh
00:11:39
Power stages I mean unless they're
00:11:42
multis sourcing the power stages for
00:11:43
this card which is possible Right like
00:11:46
we have like recently I did a video
00:11:48
about like hey you can actually get 3080
00:11:49
TI with like both uh FD MF 317s and also
00:11:54
power stages from Monolithic Power
00:11:55
Systems but I would assume like I like I
00:12:00
kind of doubt that Asus would have been
00:12:01
multis sourcing power stages for like
00:12:03
what I'm assuming is the first batch of
00:12:06
509s um so it's probably on these MP 86
00:12:12
uh uh 670 power stages um and these have
00:12:17
been used like Asus has used these on
00:12:19
like a bunch of motherboards off the top
00:12:20
of my head and I think this is the first
00:12:23
time I've ever heard of one of these
00:12:24
dying uh so I don't think this is like
00:12:27
like unless Monolithic Power systems had
00:12:29
like a defective batch I wouldn't blame
00:12:32
the power stage for this
00:12:34
um right like yeah like maybe maybe
00:12:37
there was like one one that was wonky or
00:12:40
something but I don't like this is not
00:12:42
some like bargain bucket you know power
00:12:45
stage so I don't see a reason that that
00:12:47
should fail in this application for no
00:12:49
apparent reason especially when just
00:12:51
browsing websites um now uh
00:12:57
alternatively um it could have been
00:12:59
soldered badly to the board so a concern
00:13:01
with soldering power stages is that if
00:13:03
there's too many voids in the solder cuz
00:13:05
the solder contracts as it cools um
00:13:08
which basically means it you get like
00:13:11
pockets of nothingness in inside solder
00:13:13
joints under uh surface mount components
00:13:16
with big uh P like connection pads which
00:13:19
these have the these are what the pads
00:13:21
look like so yeah that is a concern with
00:13:24
soldering of these is that you could
00:13:26
have like big gaps in the connection
00:13:28
under the power stage and that will make
00:13:30
the power stage run incredibly hot um
00:13:34
but we'd basically need like without
00:13:36
having the card in hand that's
00:13:37
impossible to to diagnose
00:13:41
um so yeah alternatively there could be
00:13:43
something wrong with the control
00:13:45
circuitry of the power stage so which
00:13:47
could be like an actual design flaw of
00:13:49
the
00:13:50
card um so it could be like running
00:13:53
wrong and then like maybe there's like
00:13:55
an issue with switching from idle to
00:13:58
high power or from you know like there
00:14:00
could be some kind of power management
00:14:01
issue or something like that um but I
00:14:05
doubt it that seems like something that
00:14:07
you you would hope that they would you
00:14:10
know catch in the development process
00:14:12
then again Nvidia has
00:14:15
been uh apparently like the 50 Series
00:14:18
development process was kind of rushed
00:14:20
but I still don't think like gpus have
00:14:23
had idle power States for years I can't
00:14:25
imagine that like the 90 has some
00:14:27
Bizarro power stage man power State
00:14:29
Management that would suddenly cause
00:14:32
power stage failures that just doesn't
00:14:33
make sense to me um so yeah so my my
00:14:39
like default assumption in this case
00:14:41
would be like oh maybe like bad
00:14:42
soldering quality on one of the power
00:14:44
stages um the main thing is or well main
00:14:49
thing like the owner of this card they
00:14:52
should just get this card replaced by
00:14:53
Asus cuz this this is not their fault
00:14:56
there's this is just like Asus sent them
00:14:58
a card with the dodgy power stage either
00:15:00
because of like soldering quality or
00:15:04
maybe the power stage itself was dodgy
00:15:06
like something was evidently wrong with
00:15:08
it and it G like you know shorted out
00:15:11
and died um ASUS ASUS should replace the
00:15:15
card um but the thing is if this card
00:15:19
had fuses it wouldn't produce fire and
00:15:22
it wouldn't produce smoke right because
00:15:25
this initial like shutdown is like this
00:15:27
is when I think the power stage like the
00:15:29
high side mosfet died which caused the
00:15:31
system to shut down and then uh with the
00:15:34
still shorted high side mosfet the owner
00:15:36
you know power cycled the system and
00:15:38
that's what created all the fire and
00:15:40
smoke because they have a 1200 watt
00:15:41
power supply because of course they do
00:15:44
they have a 590 so so you know uh if we
00:15:47
scroll through the comments yeah they
00:15:49
have a Rog Thor 1200 wat PSU if I'm not
00:15:52
mistaken this is like a relabeled seic
00:15:54
thing or something um and so off the top
00:15:58
of my head this is is going to have an
00:15:59
overcurrent protection of over 100 amps
00:16:02
right so when a high side mosfet shorts
00:16:05
out it literally just connects the input
00:16:07
12 volts to the output now under normal
00:16:11
circumstances the way the power Stage
00:16:13
Works is that it'll be this is a 90
00:16:16
probably running 500 khz so the power
00:16:19
stage is basically using that high side
00:16:21
MOS fat to uh chop 12 volts into very
00:16:25
short pulses of 12 volts into the
00:16:28
inductor
00:16:29
at like 500,000 times a second right so
00:16:32
you get bursts of 12 volts getting
00:16:34
applied to this side of the inductor
00:16:36
500,000 times a second and through
00:16:38
inductor magic which is what we're going
00:16:40
to call it for now that like those short
00:16:43
bursts or short pulses of 12 volts get
00:16:46
turned into a much lower voltage on the
00:16:49
output side of the
00:16:50
inductor now the problem is when your
00:16:53
high side mosfet shorts shorts out it no
00:16:58
longer chop up the 12 Vols right instead
00:17:00
of getting short little tiny little
00:17:02
pulses of 12 Vols CU this is you know
00:17:05
like say an 8% pulse 500,000 times a
00:17:09
second so tiny little pulses uh yeah
00:17:12
instead of getting that you just get 12
00:17:14
volts rammed straight into the
00:17:16
inductor which means the output voltage
00:17:19
starts Rising
00:17:21
uncontrollably which this is the msvd
00:17:23
rail there's not going to be a ton of
00:17:25
capacitance relative to say VC Cor
00:17:28
there's a decent chance that the core on
00:17:30
this GPU is dead now I'm sure Asus is
00:17:33
going to be thrilled about having to pay
00:17:35
Nvidia for more GPU cores because they
00:17:37
couldn't be bothered to put a actually
00:17:40
to in defense of Asus you can't really
00:17:42
put any fuses on a 5090 because the
00:17:45
power distribution on this is just not
00:17:47
designed with any kind of safety
00:17:49
features in mind
00:17:52
um like it's like the power distribution
00:17:54
on a 5090 and a 4090 as well for that
00:17:58
matter is basically designed around the
00:18:00
assumption that things don't go wrong um
00:18:04
and when things do go wrong you get fire
00:18:06
and smoke spitting out of the card um
00:18:09
because yeah you get like the entire
00:18:10
output of the power supply getting
00:18:12
dumped through this power stage through
00:18:14
that inductor and through the M like the
00:18:17
through part of the GPU core
00:18:20
um so yeah because basically like your
00:18:24
power supply like the 1200 a 1200 watt
00:18:27
power supply right when when it sees
00:18:29
that shorted out power stage if that
00:18:31
shorted out power stage is not low
00:18:33
enough resistance to immediately trip
00:18:35
the short circuit protection it will
00:18:37
just dump current into it which is how
00:18:40
you get all of the fire and smoke and
00:18:42
also probably a dead GPU in this case um
00:18:47
and the thing is if there was a fuse in
00:18:48
front of that power
00:18:50
stage in you know like that first
00:18:53
failure would have hopefully taken the
00:18:55
fuse out and then when the owner goes to
00:18:58
to turn the card on again you don't get
00:19:00
any fire because you can't apply 12
00:19:02
volts to a dead power stage a second
00:19:04
time
00:19:07
um but
00:19:09
um
00:19:11
yeah um the main point of this video was
00:19:14
though please don't blame capacitors for
00:19:16
power stage
00:19:18
failures capacitors are really really
00:19:21
reliable power stages on the other hand
00:19:24
even high-end ones aren't like I have a
00:19:26
stxs 2080 TI with a Texas Instruments
00:19:30
like 80 amp power
00:19:32
stage and I had to replace it because
00:19:35
one of them blew up I don't know why it
00:19:36
blew up but it did I have like a he I
00:19:39
have a whole bunch of 980 TI that had
00:19:41
dead power stages I just like there's
00:19:44
that's the I repaired two different 309s
00:19:46
that had dead power stages admittedly
00:19:48
those were like cheap power stages not
00:19:50
high-end ones but like power stages
00:19:52
failing is a thing that happens a lot so
00:19:55
like maybe we should design the ex Uber
00:19:59
expensive gpus to resist power stage
00:20:02
failures a bit better than just hoping
00:20:04
that the power stages don't die cuz they
00:20:07
do die a
00:20:09
lot um anyway this is almost certainly
00:20:13
not a capacitor failure um and that's
00:20:16
that's it for the video so hopefully you
00:20:18
found this somewhat interesting if not
00:20:19
particularly useful and that's it so
00:20:24
thanks for watching and oh if you'd like
00:20:27
to support the channel there are links
00:20:28
in the description you could check those
00:20:30
out it would be very much
00:20:31
appreciated and uh am I forgetting
00:20:35
anything I don't think so so I guess
00:20:36
I'll just hit the stop button now thanks
00:20:38
for watching and goodbye
