Video Fail - New Type For Me

[John McGaw]

I've been building, using, and fixing computers for longer than dirt has
been around. I finally diagnosed a new (to me) failure mode in my elderly
ATI HD 5770 that had been slowly annoying me to death. What would happen,
intermittently, was that what I'll call confetti would appear on the
screen. Most of the time it was just one or two stray dots that would
disappear if I dragged the object containing them, thus forcing a refresh.
Over what must be a year I ran every diagnostic I could find, reseated
cards, checked power supply voltages, and re-installed drivers. Finally the
problem got much worse when I would try to play back a video other than
Flash in any application I had installed. Eventually it got to the point
that a hard reboot was needed to recover most of the time.

I finally 'pulled my head out' and got a new ATI video card to test with
and the problems went away. It just goes to show that even someone with
decades of experience can overlook the obvious -- video cards don't just
die, they get sick too.

 

[Paul]

I've been building, using, and fixing computers for longer than dirt has
been around. I finally diagnosed a new (to me) failure mode in my
elderly ATI HD 5770 that had been slowly annoying me to death. What
would happen, intermittently, was that what I'll call confetti would
appear on the screen. Most of the time it was just one or two stray dots
that would disappear if I dragged the object containing them, thus
forcing a refresh. Over what must be a year I ran every diagnostic I
could find, reseated cards, checked power supply voltages, and
re-installed drivers. Finally the problem got much worse when I would
try to play back a video other than Flash in any application I had
installed. Eventually it got to the point that a hard reboot was needed
to recover most of the time.

I finally 'pulled my head out' and got a new ATI video card to test with
and the problems went away. It just goes to show that even someone with
decades of experience can overlook the obvious -- video cards don't just
die, they get sick too.

At one time, with the video cards running 90-100C, they
were calculating that the GPU would fail around the 1 year mark.
It didn't happen, but that's what the calculation showed.
(And it's not the first time that calculation was wrong, either.)
I think one of the video companies did the calc, to cast aspersions on a
competing product.

Some GPUs have left the factory, relatively poorly tested. There
were some NVidia 61xx series that there seemed to be a few too many
issues when the product was new. And it didn't seem to be
driver issues as such, as some boards got a "good one" and
others didn't

In your case, this might not have been a classical failure
as such. Perhaps it was marginal timing, and bumping VCore
or improving cooling would have been enough to fix it.

It's too bad products like that, don't come with their own
"memtest", so you can check for failures of the onboard
memory. As that will happen occasionally as well. Some
card designs, the heatsink "insulates" the memory chip
underneath it (no contact), and then the memory runs
at high temperature all its life.

Paul

 

[John McGaw]

At one time, with the video cards running 90-100C, they
were calculating that the GPU would fail around the 1 year mark.
It didn't happen, but that's what the calculation showed.
(And it's not the first time that calculation was wrong, either.)
I think one of the video companies did the calc, to cast aspersions on a
competing product.

Some GPUs have left the factory, relatively poorly tested. There
were some NVidia 61xx series that there seemed to be a few too many
issues when the product was new. And it didn't seem to be
driver issues as such, as some boards got a "good one" and
others didn't

In your case, this might not have been a classical failure
as such. Perhaps it was marginal timing, and bumping VCore
or improving cooling would have been enough to fix it.

It's too bad products like that, don't come with their own
"memtest", so you can check for failures of the onboard
memory. As that will happen occasionally as well. Some
card designs, the heatsink "insulates" the memory chip
underneath it (no contact), and then the memory runs
at high temperature all its life.

Paul

In this case, I'm just happy to keep my old daily driver working for a
while more for the modest price of $70. I don't really know if the new card
(a Radeon HD 5450) is as fast as the old card. With passive cooling it
certainly shouldn't be. My big requirement was a card with dual DVI
connectors and they've become increasingly rare. Does the job for me and no
need to mess with the drivers.

 

[VanguardLH]

ATI HD 5770 intermittently had confetti appear on the
screen. Most of the time it was just one or two stray dots that would
disappear if I dragged the object containing them, thus forcing a refresh.
Over what must be a year I ran every diagnostic I could find, reseated
cards, checked power supply voltages, and re-installed drivers. Finally the
problem got much worse when I would try to play back a video other than
Flash in any application I had installed. Eventually it got to the point
that a hard reboot was needed to recover most of the time.

I finally 'pulled my head out' and got a new ATI video card to test with
and the problems went away. It just goes to show that even someone with
decades of experience can overlook the obvious -- video cards don't just
die, they get sick too.

The quality and length of the video cable can produce artifacts, too.

Did the old video card have 1 4-pin, 1 8-pin, or 2 4-pin connections to
the PSU? What does the new video card use? What's the difference in
wattage consumption for both old and new cards? If the new video card
has fewer power connections and/or is rated at lower wattage then maybe
it's PSU going bad. PSUs wane in power output over time. I forget when
the dropoff begins but recall typical (i.e., cheap) PSUs drop 30% each
year after some point (better quality PSUs still drop but not as much)
so eventually the PSU cannot handle the existing load even if you bought
one with lots of extra capacity.

You didn't how old is the computer and if it's the original PSU or has
been replaced since and, if so, how long ago.

 

[John McGaw]

The quality and length of the video cable can produce artifacts, too.

Did the old video card have 1 4-pin, 1 8-pin, or 2 4-pin connections to
the PSU? What does the new video card use? What's the difference in
wattage consumption for both old and new cards? If the new video card
has fewer power connections and/or is rated at lower wattage then maybe
it's PSU going bad. PSUs wane in power output over time. I forget when
the dropoff begins but recall typical (i.e., cheap) PSUs drop 30% each
year after some point (better quality PSUs still drop but not as much)
so eventually the PSU cannot handle the existing load even if you bought
one with lots of extra capacity.

You didn't how old is the computer and if it's the original PSU or has
been replaced since and, if so, how long ago.

The computer is a Shuttle bare bones, X58 vintage but the power supply
voltages were checked carefully load vs no-load and the droop and noise
were excellent. If nothing else, Shuttle puts out excellent power supplies
in their little machines and for the price they charge they had better.

No, this problem was definitely a matter of the video card's guts failing
-- probably just the wear and tear of running 24X7 for years on end in a
tiny case which is not the most cooling friendly. I did the Windows
Experience benchmark today and the new card yields about two-thirds the
rating of the old one but that is fine for me -- I'm not a gamer and when
the BOINC client wants to run on the GPU I'm willing to let it wait a bit
to finish.

 

[philo]

<snip>


"Confetti" is a pretty sure sign of a video card problem.

I'm fortunate that I literally have boxes of spare parts...so as soon as
I suspect something, I substitute.