Guys - I really think I have a bad problem developing here - GTX280

[TX-Gunslinger]

For about two weeks (as far as I can remember), I've been experiencing lock-ups with my GTX-280 that I'm pretty sure have nothing to do with drivers, applications or heat.

Every so often - sometimes more often than not - when playing anything (ROF-Il2-WOP), my system freezes and the screen looks like this (pic taken from my Iphone - as I can't screen capture since the system is locked):



1. My GPU temps never get hotter than 72C under the heaviest loads I can put on it from my applications, PCB is much lower than that value.
2. I've tried various drivers - using Driver.Cleaner.net to remove all Nvidea cabs/drivers before installing new Drivers
3. I've gone back to 185.85 on my Vista partition and still same issue
4. I've experienced the same issue in XP with newest drivers
5. I've rechecked my P/S connections (modular) - and even changed the connections from PS to video card (both of them) so that the video card electrical load was distributed across both rails.


As near as I can tell - it's either the vid card or the PSU

What do you think?

S~

Gunny

[TX-Gunslinger]

The other issue I seem to have - which has made me suspect the PSU - is that I can't seem to have too many USB items plugged in (back panel), before one or two of them don't work and I have to remove something or plug it into a different port after booting to get the detected and operable by windows.

May be related, maybe not....

Sometimes, I can have everything plugged in and it's ok - and other times not.

S~

Gunny

[GOZR]

185.85  i'm running 185.52 x64

OK here make sure to test again with your PC stuck speed.. first. ( due to OC or not enough voltage ) Make sure your PCIE's GPU(s) are locked with regular speed and try again .. If you have 750 W or more your are way OK.
Make your GPU fan at 100 % speed and locked. ( it seems like GPU is dying OR your heatsink do not touch correctly the GPU or the Vram and overheat it happen to my old 6800 GT but i did fix it replacing the thermal paste on GPU and memory modules )

PSU for GTX 280's
http://www.evga.com/forumsarchive/tm.asp?m=100483867


if you have an other PSU or Video card change and test. GPU first.

Partition drives are always slower and trouble some for 80 $ get a separate one with Windows 7 x64 on it and junk Vista.

Give news plz

[TX-Gunslinger]

All this is occuring at stock clocks - after I started to have the problem.

Thanks for all the other data - will do and will check back when I"ve run those to ground....

PSU is a TOPOWER 900W....

S~

Gunny

[GOZR]

Gunny 900 W is way enough hopefully it does work ok.. check the warranty of your GPU if it is ok check the void warranty problems first before all.. and in case just do a RMA quick.. EVGA are fast. i did it for my two GTX 280's..

But to me the image you posted looks like you have to open it and check and replace thermal paste or just RMA if warranty problem.. Plz look into EVGA forum and site for more ..

Sorry gunny for this trouble it's quite frustrating and you dont need this at the moment.

[GOZR]

Hoo and make sure the GPU is well in it's socket and clean.

[TX-Gunslinger]

Gunny 900 W is way enough hopefully it does work ok.. check the warranty of your GPU if it is ok check the void warranty problems first before all.. and in case just do a RMA quick.. EVGA are fast. i did it for my two GTX 280's..

But to me the image you posted looks like you have to open it and check and replace thermal paste or just RMA if warranty problem.. Plz look into EVGA forum and site for more ..

Sorry gunny for this trouble it's quite frustrating and you dont need this at the moment.

Makes much sense.  I'm sure I registered it at EVGA.

I spent some time last night researching - and I saw where some folks had done what you suggest - checking the thermal paste and physical coupling of the VRM/GPU/RAM...

I was hoping to aviod that - but I'll pull it and check it now.

I was using EVGA Precision and GUP-z to monitor timeps and fan speed.

Thanks More later....

S~

Gunny

[TX-Gunslinger]

Those links helped tremendously.

When I think about it - this started on the the last Nvidea driver update back in Dec.  At that time I did reinstall drivers - but in following another thread over at the EVGA forums I found I'm missing a step in how I'm using Driver Cleaner Pro - i.e. in Safe Mode........

Let me go back and do that part exactly - and target drivers 185.52 (good stable driver version helps loads)

Let's see if it's the software Before I attack the hardware....

S~

Gunny

[TX-Gunslinger]

Ok - I finally beat Windows update to the punch and got a good install of certified drivers - windows update is off now for good.  Let's see what happens now.

Biggest difference in the procedure I just used was to uninstall ONLY the Nvidea Driver software and nothing else (like PhysX and the other Nvidea spam).  I hate Nvidea, btw - about as much as I hate Microsoft (have to blow some steam off every once in a while).  My greatest prayer will be answered the day ATI returns......

Now, upon reboot I F8'd and ran Driver Cleaner in Safe mode - ONLY selecting "Nvidea" versus multiple Nvidea related cleaning filters.

Upon reboot out of safe mode (which takes forever - btw - just in case your using this) - everythings up.

Gozr - the driver version your using does not show up in the archive beta list nor the certified list.... sure that number is correct?

Right now, I'm using the EVGA recommended - certified current version.  I know, I know - however I have never had a problem like this with any other 190 series drivers - so let's see where this goes....

Next step if this comes back is to try 180 series.....

Then hardware...

S~

Gunny

[GOZR]

Ho yeas ! 185.50  not 52.. was made for Vista 64 with option W7.. the newer driver do not have yet the Split SLI rendering.. but  i think the new one are good.. you should try.
The 185.50 so far is the best for me. and you can disable physx btw.

lets me know what's up//

[TX-Gunslinger]

Gunny 900 W is way enough hopefully it does work ok.. check the warranty of your GPU if it is ok check the void warranty problems first before all.. and in case just do a RMA quick.. EVGA are fast. i did it for my two GTX 280's..

But to me the image you posted looks like you have to open it and check and replace thermal paste or just RMA if warranty problem.. Plz look into EVGA forum and site for more ..

Sorry gunny for this trouble it's quite frustrating and you dont need this at the moment.

Good, good job Gozr.  My entire problem it seems was the breakdown of one of the two-bit thermal pads - in this case, the one over the NVIO chip.  More to follow, but for now......  Thanks!!! Great diagnosis - you were most certainly correct.....


Let me show you where I've been all week

S~

Gunny

[GOZR]

  thx  yes

[TX-Gunslinger]

So, after researching a little bit - I found the GPU-Z utility.  I wished I had d/l this application when I first built my system.  Like many folks, I assumed that the GPU temperature and PCB temperature that some utilities (EVGA Precision/RealTemp) monitor, was the whole story - but not at all.  There are many thermal sensors on these cards, and it just so happens that the one which brought my system to it's knees was VDDC Phase 1 thorugh 5 temperatures.

After installing GPU-Z and ATITool (ATITool is great testing utility which produces a 3D rotating cube of what looks like carpet), I saw my VDDC Phase 2 temperature skyrocket to 171.5C prior to a lockup.   Yes you read that correctly 171.5C

After reboot I selected an 80% GPU Load test (Rise of Flight - test track - max graphics).  During this test, GPU-Z revealed:


Is this hot enough, do you think?  And, yes - this is all at stock clocks on the graphics board.

I removed the card and opened it up.  On the cooler side I found:

[TX-Gunslinger]

Compare the picture in the post above, from my card's cooler to this, from a GTX-285 in perfect shape:


So let's examine the chip (NVISO) to which the broken thermal pad is connected ..


So clearly, we need to clean it up, and fix thermal material. I used an ArtiClean kit, and a LOT of swabs.  Cleaned up the whole card, including the "Beetle Dung" on the GPU.


After I removed the old Thermal Grease from the GPU, I found a small piece of tape stuck to the GPU - originating from the manufacturing process.  No thermal bond there!

[TX-Gunslinger]

At this point, I used the "Thermal Surface Purifier" from the Articlean Kit to wipe down all surfaces which I'd cleaned with the "Thermal Material Remover", which is a mild citrus-based solvent that does a great job of helping to remove Thermal Grease.  After placing the thermal pads back - I conducted some more testing.



Compared to the initial - great progress..... (Initial test repeated below for ease of comparison)



S~

Gunny