Trying to figure out a gfx card failure

[Ehud Shapira]

Hello all,


I'm trying to find out what caused my gfx card (a, well... Voodoo 3)
to go bad. I'm worried that this may affect other parts of the
computer. If anyone experienced something similar, or have an idea,
I'd love to know.

And here's the story...


VOODOO

I was using the computer when, all of the sudden, *screen corruption*
and it got stuck. Odd... I thought, then rebooted. At the BIOS
startup screens, surprise, the textmode font is all scrambled, along
with some bad colors (such as random flashing colors in the BIOS setup
screens).

That certainly didn't look good, so I turned off the computer off, and
checked the gfx card. It was rather hot, almost too hot to touch, but
I think that was still normal. I let it cool down, and in the
meantime tried another gfx card -- which worked ok.

When the Voodoo 3000 cooled I tried it again, with the same result as
earlier. I also tried it on another computer, and again, corruption
(but strangely the text font was more coherent, although still clearly
bad).


CHANCE?

My suspicion was further aroused because I had two other strange
occurances recently -- a few hours before the gfx card problem, I got
a BSOD on some networking VXD, which never happened before, but
perhaps that was related to the Realtek NIC drivers being updated
(5.05 -> 6.15) a day earlier.

Another strangness was a shutdown problem a few days earlier. After
Windows (98SE) shut down the harddrives (click), it didn't shut the
computer completely. Okay, this happens, but as I reached to turn it
off manually the harddrives sprung back on! Yuck. (I'm very touchy
when it comes to harddrives). First time I see this.

It's quite possible that this is all just chance, but maybe there's a
more fundamental problem which may yet wreak more havoc?


SUMMARY

* Anyone ever experienced something similar with a gfx card?
* Anyone ever had the harddrives turn back on after Windows turns them
off at shutdown?
* Anyone ever had a BSOD related to networking? (nah, forget that).

So what are your thoughts? Should I be further concerned, or is this
just a gfx card gone bad?


Ehud.

 

[kony]

Hello all,


I'm trying to find out what caused my gfx card (a, well... Voodoo 3)
to go bad. I'm worried that this may affect other parts of the
computer. If anyone experienced something similar, or have an idea,
I'd love to know.

And here's the story...


VOODOO

I was using the computer when, all of the sudden, *screen corruption*
and it got stuck. Odd... I thought, then rebooted. At the BIOS
startup screens, surprise, the textmode font is all scrambled, along
with some bad colors (such as random flashing colors in the BIOS setup
screens).

That certainly didn't look good, so I turned off the computer off, and
checked the gfx card. It was rather hot, almost too hot to touch, but
I think that was still normal. I let it cool down, and in the
meantime tried another gfx card -- which worked ok.

What happens if you boot all the way to (windows?), or at least
try to?

When the Voodoo 3000 cooled I tried it again, with the same result as
earlier. I also tried it on another computer, and again, corruption
(but strangely the text font was more coherent, although still clearly
bad).

This Voodoo, does it use a fan and is that fan operational?
If it'd always ran hot, could simply be the long-term effects of
that... heat is certainly a limiter on hardware lifespan,
especially when hardware has electrolytic capacitors on it, but
I'd thought Voodoo 3 typically wasn't wearing any (electrolytics)
except a few tiny ones around memory power circuit. If text is
consistently corrupt it sounds more like a video bios corruption.
I presume the fonts look just as bad when card is cold, from
first power-on after system had been sitting off, as when left on
or retried for a few minutes.

CHANCE?

My suspicion was further aroused because I had two other strange
occurances recently -- a few hours before the gfx card problem, I got
a BSOD on some networking VXD, which never happened before, but
perhaps that was related to the Realtek NIC drivers being updated
(5.05 -> 6.15) a day earlier.

Could be... if problem persists just revent back to old driver.
Was network driver a problem previously or was there no technical
motivation to update it?

Another strangness was a shutdown problem a few days earlier. After
Windows (98SE) shut down the harddrives (click), it didn't shut the
computer completely. Okay, this happens, but as I reached to turn it
off manually the harddrives sprung back on! Yuck. (I'm very touchy
when it comes to harddrives). First time I see this.

It does sound odd but perhaps one or more drives were asleep?
WIndows may spin-up drives that are asleep when it shuts down.

It's quite possible that this is all just chance, but maybe there's a
more fundamental problem which may yet wreak more havoc?

Well the key to isolating this may be determining if these
problems still occur after:

a) reinstall of old network driver
b) use of different video card

SUMMARY

* Anyone ever experienced something similar with a gfx card?

Sure, they all die in one way or another, except for the old
worthless 1-2MB PCI cards, those seem to live forever.

* Anyone ever had the harddrives turn back on after Windows turns them
off at shutdown?
* Anyone ever had a BSOD related to networking? (nah, forget that).

So what are your thoughts? Should I be further concerned, or is this
just a gfx card gone bad?

Best thing to do is just tackle each problem one at a time, then
see what problems remain, focusing on video card first since it
acts up even before OS loads.

You might also look for failed capacitors on video card,
motherboard, and check power supply voltages with a multimeter.

 

[John]

SUMMARY

* Anyone ever experienced something similar with a gfx card?

Yes but unfortunately Im not 100% sure what caused it. Well Im pretty
sure it was my motherboard but Im not sure what was the specific
cause.

I had a Nvidia Geforce 2 card that got super hot. My Abit was a bit
wonky - via chipset kt333 anyway. It was ultra stable then it started
going downhil in about a year. Samething happened to a variety of MBs
Ive had the last few years for different reasons.

Was it bad caps ? I looked and didnt see any of the obvious signs -
leaks or bulging. Who knows. Had other problems with ASUS too.

Anyway - my MB started locking up so I look and my fan is dead on my
card. I also notice my card is getting hot. I went through a long
trial and error thing after that trying to fix my card and after
buying another card and fan for my old card still had the problems.

I ended up buying two new 400 mx cards . Turns out my old card worked
perfectly with a new fan and its was probably my M B. I then went
through $$$ as I did the usual on several systems actually - swapping
PSes (bought two new Antecs and a Compusa) , mem, motherboards, hard
disks etc and took the MBs out of the case and etc Real pain the ass.

The Gfx card would get very hot ! And when I tried the newer Nvdia
drivers it got almost glowing hot. I kept shutting it down. As I
monitored the temps with the fingers on the card method.

Finally it died. I looked at saw a small burn mark at one of metal
contacts on the gfx socket. They sent me a new gfx card and Abit sent
me a replacement MB as I threw in the towel and finally narrowed it
down to the MB probably.

After that no such problems - but I had other problems with the
replacement board and other boards which were also replaced by Asus
and Abit.

* Anyone ever had the harddrives turn back on after Windows turns them
off at shutdown?
* Anyone ever had a BSOD related to networking? (nah, forget that).

So what are your thoughts? Should I be further concerned, or is this
just a gfx card gone bad?

How about the fan on your card. Thats an obvious thing to check. Temps
on your system ? Air flow through your system? I started using a prog
to monitor HD temps and was surprised how high things got with some
programs like quickpar. In fact I used one SMART monitoring software
and for some reason the HD light wouldnt go off after I installed it
and temps started going higher and higher. I took that thing off and
no problems since then. I use fans on all my HDs now.

If you are getting other problems could be the MB or other things
obviously from the PS to HD corruption/virus though I think youd
notice that. There was the bad cap thing going around a while back and
Ive had lots of other problems - Ive never got the Abit Highpoint
controller to not lockup intermittently on the replacement board
though my original one that melted down worked fine. A old KT133a Asus
started locking up - I got a replacement and that locked up too. I
finally traced it to a intermittently working northbridge fan.

Recently Ive had some problems that seemed to be related to corrosion
- I live in a really humid place by the beach and you should see the
insides of the cases especially around the fan in and out areas - lots
of rust . The KT133a system went nuts again - i mean totally nuts
intermittently the last 6 months again until I cleaned the mem
contacts and CMOS battery contacts.

 

[Ehud Shapira]

@ kony:

I didn't try booting to Windows. Seeing the bad textmode was scary
enough for me. Something DID happen there, and even if a BIOS
reflash may fix it, things like that shouldn't happen, so how could I
trust this card again?

This Voodoo 3 is fanless, just the way I like it, and even though it
heated I think that was normal.

The card looks fine, including capacitors, and no burn evidence.
It has those smaller capacitors (see
http://www.techspot.com/reviews/hardware/voodoo3-3000/card.jpg), I
wonder why more modern cards tend to use the classic big ones.

The network-related BSOD happened just once, and didn't really bother
me. What alerted me was the bad gfx card, which made me extra
suspicious of every small abnormality.

I don't like harddrives turning off on me (even if I'm not there), so
this feature was disabled... Windows clearly shut the drives off
first.

Could you elaborate on the cards that failed on you? Did it happen in
middle of work, or when you booted, etc.?

@ John:

Ah, the horror stories you tell! And rust?! Luckily I have none
of that.

I've had some weird occurances two years back, and I'm still not sure
what was the cause. What I finally did was get a new Enermax PSU
instead of the noname I had before.

As for the current issue, voltages seem fine (as reported by the
motherboard).
And I my motherboard and gfx are both fanless -- I really try to avoid
fans (although in this case it's just that I have old stuff, before
fans started being the sign of "high performance gear".)

Heat for motherboard/CPU is normal. Neither one of my harddrives
support temperature reporting (I've had an evil IBM DTLA which did
report temperature, but it's now dead [including its replacement], and
strangely the newer WD 400JB doesn't).

 

[kony]

@ kony:

I didn't try booting to Windows. Seeing the bad textmode was scary
enough for me. Something DID happen there, and even if a BIOS
reflash may fix it, things like that shouldn't happen, so how could I
trust this card again?

If it was merely the bios, you'd just have to try it and see.

This Voodoo 3 is fanless, just the way I like it, and even though it
heated I think that was normal.

Yes it was "normal", but, normal because it was just designed to
be servicable for a fair period of time, at this point it's a 6
year old card.

The card looks fine, including capacitors, and no burn evidence.
It has those smaller capacitors (see
http://www.techspot.com/reviews/hardware/voodoo3-3000/card.jpg), I
wonder why more modern cards tend to use the classic big ones.

They are for the memory, which being SDRAM, can use 3.3V direct
from AGP bus. Newer cards have higher power requirements, but
more significantly they typicially have lower voltage from a
switching regulator, which in itself requires large caps. Often
on those newer designs they do still have the smaller caps too,
performing same function as on the Voodoo.

The network-related BSOD happened just once, and didn't really bother
me. What alerted me was the bad gfx card, which made me extra
suspicious of every small abnormality.

I don't like harddrives turning off on me (even if I'm not there), so
this feature was disabled... Windows clearly shut the drives off
first.

Maybe it's enabled in the bios?

Could you elaborate on the cards that failed on you? Did it happen in
middle of work, or when you booted, etc.?

More often I get them DOA, unknown exactly when they failed. I'd
expect most that failed were working till system was powered
off/on again, or locked up system and then it was reset.
Sometimes corrupt font like yours, sometimes no picture at all,
sometimes no picture till windows, sometimes motherboard won't
turn on, other times banding or smearing of contrast or
snowy/speckled look... many different signs.

With older parts it may not be worth worrying about why it died,
just getting the system operational again takes priority then
futher research is warranted if there are still problems.

 

[Ehud Shapira]

If it was merely the bios, you'd just have to try it and see.

Who said next time won't be more disastrous?

Yes it was "normal", but, normal because it was just designed to
be servicable for a fair period of time, at this point it's a 6
year old card.

I think fans for low-end cards today are just to entice unknowing
customers. A good heatsink should do just as well as a bad heatsink +
fan, but sans the unavoidable failure. Those passive Sapphire
9600/9800XTs are my ideal.

I wish companies (nV/ATI/Intel/AMD) started caring more about heat
issues, and less about squeezing every MHz. How nice it'd be if in 3
years active cooling would be a thing of the past? I hope future
Intels will share more with Pentium M and not Pentium 4.

Maybe it's enabled in the bios?

Nah, no sleeping at all, but even if it was, it's unrelated. The
harddrives thing was perfectly odd.

More often I get them DOA, unknown exactly when they failed.

Where do you get all those DOAs?

I'd expect most that failed were working till system was powered
off/on again,

That's the scenario I would find more plausible, and not what happened
here.

With older parts it may not be worth worrying about why it died,
just getting the system operational again takes priority then
futher research is warranted if there are still problems.

System is operational. I'm at the post-traumatic preemptive analysis
stage, as
I'm worried of potential catastrophic hard-drive failures -- all too
many HDDs died on me, and even although most weren't complete
failures, my hard-drive paranoia level is always high.

I'll have to have complete redundancy one day... that'd be nice.

(--END--)

 

[kony]

Who said next time won't be more disastrous?

Who said a different video card wouldn't fail even faster?
IF you got it working again, and problem was bios, it could
easily continue working fine.

There are no "sure things", all parts could fail, even new ones.

I think fans for low-end cards today are just to entice unknowing
customers. A good heatsink should do just as well as a bad heatsink +
fan, but sans the unavoidable failure. Those passive Sapphire
9600/9800XTs are my ideal.

No, a good heatsink without fan is not as good as small fanned
heatsink, unless small fanned heatsink was VERY poor. It is well
known that your choice for fan reduction is not "free", component
life does get shorter, though it'snot to suggest everyone needs
high-RPM LOUD fans. The Sapphire cards will also have shorter
lifespan without fan. Realize that there is a difference between
"cool enough to be stable at the moment" and "cool enough to
promote long lifespan".

I wish companies (nV/ATI/Intel/AMD) started caring more about heat
issues, and less about squeezing every MHz. How nice it'd be if in 3
years active cooling would be a thing of the past? I hope future
Intels will share more with Pentium M and not Pentium 4.

Intel is already going that route, but if you really care more
about heat you can underclock and undervolt products yourself,
it's particularly easy with CPUs and many video cards relative to
other parts.

Nah, no sleeping at all, but even if it was, it's unrelated. The
harddrives thing was perfectly odd.



Where do you get all those DOAs?

Here, there, everywhere. Years ago worked at local PC shop,
still consult for them and do a few repairs in spare time, plus I
supply parts to same and other local shops and can often get
stuff given to me or at least offered for so little $ it's more
like a storage fee for them to keep it lying around till I come
and get it, rather than a value per parts. Not just dead video
cards though, usually most of a system(s) that have
who-knows-what inside, maybe a part failed or maybe not...
sometimes all that was wrong was that they installed AOL, LOL,
and the shops bench fees were so high they opted to just replace
system instead of having it redone... any low-value tech can pop
parts together and install WinXP but those that can "fix" are
often in shorter supply, their time more valuable, so a customer
looking at over $100 fees for a 2+ year old system may often
choose to just buy a new one.

That's the scenario I would find more plausible, and not what happened
here.



System is operational. I'm at the post-traumatic preemptive analysis
stage, as
I'm worried of potential catastrophic hard-drive failures -- all too
many HDDs died on me, and even although most weren't complete
failures, my hard-drive paranoia level is always high.

I'll have to have complete redundancy one day... that'd be nice.

Not very hard to do unless you work with massive video file
collections, since today a 200GB drive can be had for just over
$100, and with more and more motherboards having RAID features.

If you see a bunch of parts failing it's often a power supply
problem.

 

[Ehud Shapira]

There are no "sure things", all parts could fail, even new ones.
I'd rather trust new hardware than something that's failed before.

But... to a different topic.

Can you e-mail me, so that I could contact you directly? I'm trying
to find someplace/someone to buy from a certain piece of (used)
hardware, and if you have all kinds of scrap hardware in transit
there, maybe you can help me.
(No luck at eBay).


Ehud.

 

[emmanuel poirier]

Hi !

Your video card is dead and especially its rams !
If it has happened this is perhaps due to too much heat as you said
others parts have problems. But W98SE is bad and drivers are also a
cause of crashes, try to update both of them.