ACK! I think 3DMark killed my card!!

[BD]

Hi, all.

I have a 9800Pro 128MB from a couple/three years ago.

I recently bought a refurb'd X850Pro direct from ATI.

I wanted to get a comparison of speed before and after the change, so I
ran 3DMark06 against my 9800.

I didn't worry much about changing settings or selecting tests; I
figured that if the card couldn't handle a certain function, then the
test would not be run.

I saw a whole mess of graphical corruption on the screen after a while.
And now, I see the same corruption in games, and in normal windows.

I'm curious if there's a possibility the 3DMark tests might have smoked
my card's memory or something? I had something similar happen to a
GeForce some years ago; I started seeing stray polygons in some
Direct3D Games; ended up pitching the card.

I can see the corruption as I type this, and it was captured with a
PrtScr.

I've uploaded a sample image to
http://members.shaw.ca/robertrd/corruption.jpg, in case anyone's
curious.

Interestingly, in the time that it's taken to draft this post, the
corruption has dissipated. I just came out of a game in which it was
rather severe... Evidence suggests that it starts to manifest as
intense graphics processing starts (ie a game) and subsides afterwards.

NEVER seen this before, and I've had this system unchanged for about 3
years.

Any opinions?

Thanks!!

BD

 

[Andrew MacPherson]

http://members.shaw.ca/robertrd/corruption.jpg,

Want to try that again?

Sounds like your refurbished card wasn't tested properly to me. 3dmark06
probably pushed every last little bit of the card harder than most
current games ever will.

Andrew McP

 

[Andrew MacPherson]

Sounds like your refurbished card wasn't tested properly to me.

Forgot to say the first thing I'd try is taking the side of the computer
off and pointing a desk fan at the internals. The first problem to rule
out is always heat (for me anyway).

Andrew McP

 

[BD]

Sounds like your refurbished card wasn't tested properly to me.

The current one (9800) was new, not refurb. The X850 which is on the
way is refurbd.

I'll check with ATI before I run 3DMark on the 850. Or maybe it's not
worth doing... all it will give me is a number. If refresh is faster,
that's the main thing.

 

[BD]

The first problem to rule

out is always heat (for me anyway).

I also considered letting the machine stay off overnight.

I never had this corruption before today, so if the whole thing is too
hot, it's unlikely that it's too hot _only_ as of today.

Heat's probably the issue, though, given the evidence.

 

[First of One]

I'm curious if there's a possibility the 3DMark tests might have smoked
my card's memory or something? I had something similar happen to a
GeForce some years ago; I started seeing stray polygons in some
Direct3D Games; ended up pitching the card.

Nah. The memory didn't have heat sinks coming out of the factory, so nothing
has changed with respect to memory cooling in the three years you've owned
the card. A software application can never "overstress" a card and cause
component failure, only uncover an existing weakness in the cooling, etc. In
any case there is no danger in running 3DMark again on your X850.

In fact, 3DMark isn't even the most efficient tool for stressing the video
card. Run the artifact tester in ATiTool. It'll put the GPU under maximum
load in a matter of seconds and pinpoint any graphics corruption down to
single-pixel artifacts.

Interestingly, in the time that it's taken to draft this post, the
corruption has dissipated. I just came out of a game in which it was
rather severe... Evidence suggests that it starts to manifest as
intense graphics processing starts (ie a game) and subsides afterwards.

So it seems like heat. It is quite common for the cooling to deteriorate
with age. Dust builds up on the heat sink. The sleeve-bearing fan loses
lubrication. The thermal paste dries and cracks...

NEVER seen this before, and I've had this system unchanged for about 3
years.

It's quite possible the card has been displaying artifacts for months, but
the visual impact has been small enough to go unnoticed.

 

[BD]

card. Run the artifact tester in ATiTool. It'll put the GPU under maximum
load in a matter of seconds and pinpoint any graphics corruption down to
single-pixel artifacts.

Will do. Thanks!

It's quite possible the card has been displaying artifacts for months, but
the visual impact has been small enough to go unnoticed.

Mmmmm... no. I'd have noticed. I'm the guy who runs Dead Pixel Buddy on
my LCD monitor every couple of months and looks for blown pixels, and
stares at the solid colors for any evidence of changes in the lighting
elements. If there were graphic aberrations, I'd have known. ;-)

 

[BD]

Run the artifact tester in ATiTool.

No errors after 1:40. I guess that's good? Never used that tool before.
It's not something I should let run for hours or anything, like a
burnin, hm?

 

[Andrew MacPherson]

The X850 which is on the
way is refurbd.

Sorry, thought you were talking about the new card. That'll teach me to
post at 5am. ;-) It's possible the 9800 is in its last legs... electron
migration etc. can eventually "wear out" a chip which runs too hot for
too long. You'd probably only notice such wear & tear when thrashing the
card.

Andrew McP

 

[Andrew MacPherson]

Run the artifact tester in ATiTool.

Good idea. I use ATItool of both my PCs (to underclock 95% of the time),
and it has a variety of uses.

Andrew McP

 

[First of One]

It's very unlikely you'll notice single-pixel artifacts *in-game*. And
sometimes it could be low-contrast like a green pixel turning yellow.

 

[First of One]

It can be used for burn-in. However, it will continue running if minor
artifacts are detected. And since there's no logging feature, you'll have to
sit at your PC for a while and watch for artifact messages. Normally,
because the GPU heats up quickly, if you see nothing for 5 minutes, then
nothing will happen.

 

[BD]

It's possible the 9800 is in its last legs...

I think that's what's happened. But it's also clear that whatever
'verge' of a problem it's been on, the 3dMark pushed it over. Must have
done, as the card's been fine for 3 years, and now it's not.

Anyway. Once this refurb'd X850 shows up I should be fine for another
while.

FWIW, I pulled the card and found a lot of dust.

But I also found that when I pulled on the heatsink, it raised up off
the chip. It seems that the spring load is the only force holding the
heatsing in place. I wonder if some conductive goop would make any
difference in this case...

 

[KCB]

I think that's what's happened. But it's also clear that whatever
'verge' of a problem it's been on, the 3dMark pushed it over. Must
have
done, as the card's been fine for 3 years, and now it's not.

Anyway. Once this refurb'd X850 shows up I should be fine for another
while.

FWIW, I pulled the card and found a lot of dust.

But I also found that when I pulled on the heatsink, it raised up off
the chip. It seems that the spring load is the only force holding the
heatsing in place. I wonder if some conductive goop would make any
difference in this case...

Just make sure you clean off the old stuff before applying the new.

 

[BD]

Just make sure you clean off the old stuff before applying the new.

No clue what's in there at this point; I certainly never applied
anything. But yeah, I'll track some of it down. I've used it before for
a CPU heatsink. I think it came in a tape kind of form.

 

[BD]

Quick follow-up:

As I was cleaning the dust from the heatsink I noticed that it was not
actually _attached_ to the chip; the spring clips were holding it down,
but that's it. Struck me as odd.

Went and bought a tube of Arctic Silver.

Took off the sink, scraped the dried crap off the chip and the sink
with a razor blade, added the new goop, and *presto*. Problem solved.

Arctic Silver is my friend.

 

[Garrot]

In fact, 3DMark isn't even the most efficient tool for stressing the video
card. Run the artifact tester in ATiTool. It'll put the GPU under maximum
load in a matter of seconds and pinpoint any graphics corruption down to
single-pixel artifacts.

I don't know about that. I ran ATI Tool and had no issues at all but
running 3DMark06 gave me screen corruption in the Deep Freeze bench towards
the end. If I lower the speed of the graphics ram I can run the same test
fine.