Need help diagnosing GPU problems

[haifisch]

Hi,

I got a Palit GTX 660 here with some problems. I know it's faulty, because when I put into my system yesterday, it caused BSOD when installing drivers and at windows startup(Wich was because of the aero theme I guess).

So I put together another system from parts I have lying around to test the card and not have to bother with my setup. I made a fresh windows install, then I could successfully install the latest nvidia drivers, but if I use windows aero, google chrome, start any game(even minesweeper!!!), or Furmark, the screen goes black in a few seconds and then restarts. The only "benchmark" I could run so far is the render test in gpu-z. I can play the sample video "Wildlife" in windows 7, no problems. More interestingly, I can run Folding@Home with no problems. It's a very GPU intensive task, so I dont understand why I cannot run Furmark, what's the difference between them?

Of course, I have the pcie power connector plugged in, and I also checked at gpu-z for temps, wich are good. Power supply is a Delta 750W.
I also made logs from these events described above with gpu-z, I attached them to this post.

Now I know I could just use another card, but I really want to fix it. My question is: Where to further test the card? Underclocking? gpu bios flash? looking for visible damage? If any of them, what programs to use? What part of my card is failing?

Thanks for any help in advance! Also let me know if any important information is missing

UPDATE:
I tried the card in my main system(with a test hdd for another os install, so I don't mess up with mine) There it doesn't restart, but screen goes black for few seconds and I get a notification on the taskbar "display driver stopped responding and has recovered". I made a video about what happens if I run the windows experiece rating, its pretty interesting, it crashes a lot of times during the test, and at the end the gtx 660 gets a score of 2,0...
Here is the video:

UPDATE #2:
I forgot to upload what error Furmark gave

 

[st2000]

not sure, but may be sensors are wrong
try to downscale gpu/memory clocks, also try to increase fan speed(as i see from sensors - it never goes over 30%)

 

[haifisch]

UPDATE

 

[haifisch]

not sure, but may be sensors are wrong
try to downscale gpu/memory clocks, also try to increase fan speed(as i see from sensors - it never goes over 30%)

Thanks for the tip, what program do you suggest to use to downscale the clocks?

 

[st2000]

use msi afterburner(tried it with my video card - 540 to 1500 mhz), but dont forget to aply results
try to downscale gpu first(i'll advice you to change gpu clock to 928 mhz and if it's stable,step by step +5mhz and test, if it's still not stable -50mhz+test)
sor for long responce time)

 

[st2000]

also searched for same problems:
"Have you tried updating Intel Graphics driver? Without the Intel Graphics drivers installed the integrated CPU graphics fails when switching adaptive power mode and sometimes randomly. Setting Adaptive power in NV control panel to Maximum Performance can also fix the problem if Adaptive power isn't working properly. Set BIOS option to PCIE 2.0 or 2nd Generation if your BIOS has it, as the Auto option or PCIE 3.0 can cause the graphics driver to not respond. May want to check the graphics card is seated comfortably into it's slot and all power connectors are secure."

 

[haifisch]

also searched for same problems:
"Have you tried updating Intel Graphics driver? Without the Intel Graphics drivers installed the integrated CPU graphics fails when switching adaptive power mode and sometimes randomly. Setting Adaptive power in NV control panel to Maximum Performance can also fix the problem if Adaptive power isn't working properly. Set BIOS option to PCIE 2.0 or 2nd Generation if your BIOS has it, as the Auto option or PCIE 3.0 can cause the graphics driver to not respond. May want to check the graphics card is seated comfortably into it's slot and all power connectors are secure."

Thank you for this
I don't have onboard video, my i5-650 has, but the mobo doesn't support IGP. I did clean windows on that drive, and installed newest nvidia drivers, also set maximum performance in nvidia control panel already. I might look into the PCIe options, but I think my board is too old for 3.0. Gonna try msi afterburner later this night and set the clocks you suggested

 

[Toothless]

Seems to be unstable at the current clock. Underclocking would help or uping the voltage might help.

 

[haifisch]

Updated with image, now I m trying to underclock/add some more voltage, but it's the first time I m using afterburner, how do I set an exact voltage/freq etc, I can only set +xx or -xx, I dont really get it :/

 

[haifisch]

I tried undervolting core, memory, and tried to add a bit more voltage while underclocked. I went all the way down to core clock -350 mhz(-50, then -100, -150,...) and memory clock -400 mhz. As for voltage, it's +75mv now(+25, +50, +75). So now I am at -350mhz on core, -400mhz on mem and +75mv on core voltage, and I still get the error with Furmark. And even minesweeper still crashes.
Should I go for reballing the chip? Is there a way to do a separate vram test, so I can know which component to reball, gpu or memory.
Do you think the VRM modules could be bad? I don't think so because I can run my full GPU power, it just depends on the application, Folding@Home works fine, no crash there.

 

[st2000]

now i can say that it's very high potential that chip is broken(mb some layers)
my last advise is to check pciE lines on card and mobo(check it and try to clean)

after searching error code 7 - try to disable gpu physx

 

[haifisch]

now i can say that it's very high potential that chip is broken(mb some layers)
my last advise is to check pciE lines on card and mobo(check it and try to clean)

after searching error code 7 - try to disable gpu physx

Tried disabling physx, but same error with furmark. I do a clean windows installation once again, and try the things you suggested here one more time. After that, I might try flashing the vbios if you think it might helps. And then if that doesn't, I take the card to a technician who has a reballing station, and try a bga reball. I'll keep updating until gpu is alive or I gave up

 

[igorm123]

Can someone help me with gpu z log file i downscale clocks on gpu and i got vertical lines with one game assassin's creed unity.... Gpu is amd radeon hd 6970

 

[haifisch]

Can someone help me with gpu z log file i downscale clocks on gpu and i got vertical lines with one game assassin's creed unity.... Gpu is amd radeon hd 6970

Are you overclocking? What kind of 6870 do you have? I mean asus gigabyte msi...etc
Can you make a picture of the artifacting?

 

[igorm123]

I didn't overcloked i downscaled memory and gpu clock's, from 1375 to 1200 Mhz and from 880 to 800... don't know what is i know is only amd hd radeon 6970 here is picture from gpu z...

 

[haifisch]

I didn't overcloked i downscaled memory and gpu clock's, from 1375 to 1200 Mhz and from 880 to 800... don't know what is i know is only amd hd radeon 6970 here is picture from gpu z...

Why are you downscaling? Try running furmark on stock clocks and see what temps and artifacting you get. Maybe your temps are a little high, idk. Upload the log file and the screen shot while running furmark if you like
If you find your card overheating, you can clean the heatsink of your card, or try changing thermal paste on the gpu if cleaning doesn't help.

 

[igorm123]

I downscaled clocks because won't work in furmark with 1375 Mhz i got vertical blue lines, i downscaled to 1200 and work perfect for 1h no overheating. This problem i have it only happend with assassin's creed unity vddc on gpu z log file jump enormly... Every other game works fine without problems or vertical lines...

 

[CAPSLOCKSTUCK]

this has become a different thread.

@igorm123 you might get more responses if you start a new thread specific to your issue

B Team

 

[igorm123]

I statrted ne thread sorry for this post's