Sign in with
Sign up | Sign in
Your question

GTX580 artifacts etc. Whats could be the possible causes?

Last response: in Graphics & Displays
Share
February 9, 2011 1:49:39 AM

I'll try and make this a long story short.
End of November built my system with a GTX580.
Everything is running sweet and no dramas.
Weekend just gone Sunday the 6th of February I turn the monitor on in the morning after leaving the PC on over night like it normally is and get confronted with massive artifacts and tripped out colours on the game that was running. No thunderstorms or advers weather that night either.
First thought is my GPU has died in some way.
I exit the game, reboot, relaunch the game and it's still the same. I try watching videos online and playing another game and the result is the same. I launch GPU-Z and the temp is well within range at 41 degrees C.
I contact my place of purchase and fill out the RMA and get that underway Sunday night the 6th.
Monday the 7th I download every GPU stress testing program I can find in preparation to go through them on Tuesday night. At this stage the problem still exists.
Tuesday night the 8th I get home from work, turn the monitor and and try running a game to reproduce the artifacts again and to my amazement it works. No artifacts or glitches of any kind. I proceed to run every program I D/l'd the day before and sift through the results.

OCCT is the only program that reported any erros with my GPU's memory. With every memory bank it tested it returned between 40 and 50 thousand errors. It didn't explain what the errors were or what they could even be, only that they were there. Keep in mind the GPU is still working fine at this stage and running games/vids etc.
I run OCCT, Furmark, Memtest, Prime95 and Intel burn test on the highest settings they have for atleast an hour each run with repeated alternate settings inbetween runs and the card holds up everytime along with all my other components.
MOBO memory reported fine, CPU reported fine, nothing over temp, GPU temps didn't peak above 74c and it still works fine.
I have been told it may be a driver glitch for the GPU and that windows 7 does auto updates on stuff like this and can sometimes currupt the drivers already installed with what windows thinks is a better driver.

I really would like to hear peoples opinions on what else I can test and with what program, or what it could have been that caused the imtermittent fault. I'm not at the pulling my hair out stage yet but I'm sure if it happens again I will be. :pt1cable: 

P.S. The long story short didn't happen...lolz

More about : gtx580 artifacts whats

a b U Graphics card
February 9, 2011 2:16:28 AM

Try reinstalling the latest drivers.

The OCCT memory test program is a CUDA app that tests the GPU memory. If it still gives errors after the newest drivers, something may be wrong with the card. However, 40-50k errors should cause problems in applications/games, *unless* the game is not using the defective memory section.
m
0
l
February 9, 2011 3:02:48 AM

I shall try that.
What could the errors actually mean though? Does it mean some of the memory is corrupted somehow or burnt out? With so many errors and nothing to tell me what they are it kind of feels like it's teasing me a bit by telling me something is wrong, but not telling me what it is.
m
0
l
Related resources
a b U Graphics card
February 9, 2011 3:07:59 AM

I imagine the code does something simple like writing a pattern into memory and reading it back out. If out doesn't match in, it is an error. Did it give a "position" (such as memory address) for the errors, and if so were they always in the same place? (Sorry for all of the questions, I only used the OCCT memcheck once, and on my card it gave no errors, so I never ran it again).
m
0
l
February 9, 2011 3:12:06 AM

I think what I shall do is run it again tonight and post the screenies up for you to check out if that's ok?
m
0
l
February 9, 2011 3:17:40 AM

Are there any other programs you know of that give more info on the GPU memory?
m
0
l
February 9, 2011 3:43:21 AM

My system is as follows;

EVGA X58 FTW3 Motherboard 132-GT-E768-KR ~ OCZ Z Series 1000W Gold Modular ~ Intel Core i7 950 ~ OCZ Vertex 2 3.5\" 240GB E Series SSD ~ 12GB Mushkin 998958 Radioactive PC3-12800 (3x2GB) DDR3 ~ Gainward GeForce GTX 580 1536MB ~ Windows 7 64Bit
m
0
l
a b U Graphics card
February 9, 2011 12:05:47 PM

GPU-Z will tell you the amount of memory being used. Otherwise, texture stressful games and benchmarks should heavily load the ram. So try the free copies of 3DMark06 and Vantage (and 3DMark11 if you'd like). Also, you could try underclocking the GPU/Mem if you still get errors. However if you keep playing high graphics games with no issues whatsoever, it makes me think there may be a bug in OCCT or the drivers causing the error rather than hardware.
m
0
l
a b U Graphics card
February 9, 2011 12:11:18 PM

Was an error file generated (in I assume the OCCT program folder)?

After a little searching, I've seen several instances where erroneous errors were generated in OCCT GPU Mem. Do you have the newest version? Even if you do, I'd be inclined to ignore its results unless crashes in games/benches back it up.
m
0
l
February 9, 2011 7:59:59 PM

I re-ran OCCT again and got not a single error for the GPU memory!!! Go figure!. I ran stress test after stress test and everything was fine. For the moment I'm just going to leave it as is and hopefully it was just a one off. If it happens again I'm going to get some vids and youtube them to show what it does. Thanks for your help mate.
m
0
l
a c 609 U Graphics card
February 9, 2011 8:39:27 PM

Download and install the latest drivers from the Nvidia.com website. Do not rely upon Windows to auto-update or fix the drivers.
m
0
l
a b U Graphics card
February 10, 2011 1:41:52 AM

Glad to hear. Hopefully just a passing software bug.
m
0
l
!