• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

GPU MemTests on GTX 480 showing weird results..

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

r2rX

Registered
Joined
Jul 2, 2011
Hey guys,

I've had an interesting week, so far, trying to figure something out. Basically, i've been having game/system crashes for the past week to two. Irrespective of operating system (Windows 7 / 8) or drivers, i've been having system hang-ups and BSOD's; mostly relating to memory-related issues.

Recently, I set up a system consisting of a Core i7 920, 6GB Corsair Dominator GT RAM, MSi Big Bang X-Power mobo, SilverStone 1KVA PSU and the GTX 480 (which I had before).

The following issue occurs with all hardware set to default settings (no OCing).

Now, the BSOD's were all memory-related errors; naturally, I tested my system RAM extensively with Memtest+..... not a single error.

What caught my attention was that if the system didn't hang-up/BSOD, the game i'd be playing would quit to Windows...and I would either get an error reporting indicating the nVidia driver or DirectX being the culprit for the crash OR (if there was no error log) i'd try run the game again...and it would tell me that my GPU does not support the "necessary API features"......?

This is when I thought to test my GPU memory. Tests occured using MemtestCL v1.0, MemtestG80 v1.1 and OCCT v3.1.0 (v4.2.0 didn't have a memtest tab).

Between MemtestG80 (nVidia/CUDA-only) and MemtestCL, i'd run 30 passes with 1344MB of GPU memory to scan (selecting 1536 would not work; the system reserved 192MB worth...so fine, 1344MB it is).

Code:
Test iteration 30 (GPU 0, 1344 MiB): 19 errors so far

        Moving Inversions (ones and zeros): 0 errors (47 ms)
        Memtest86 Walking 8-bit: 0 errors (390 ms)
        True Walking zeros (8-bit): 2 errors (187 ms)
        True Walking ones (8-bit): 0 errors (187 ms)
        Moving Inversions (random): 0 errors (63 ms)
        Memtest86 Walking zeros (32-bit): 0 errors (764 ms)
        Memtest86 Walking ones (32-bit): 0 errors (764 ms)
        Random blocks: 0 errors (344 ms)
        Memtest86 Modulo-20: 0 errors (2324 ms)
        Logic (one iteration): 0 errors (31 ms)
        Logic (4 iterations): 0 errors (47 ms)
        Logic (shared memory, one iteration): 0 errors (31 ms
        Logic (shared-memory, 4 iterations): 0 errors (63 ms)

The above is a typical results from MemtestG80. This is what confuses me.... MemtestCL has never shown a single error..... MemtestG80, however, would show me errors (with 30 passes, it would range between 10-20 errors).

When MemtestG80 would show me the errors, it was one test in particular: True Walking zeros (8-bit). If looking at the sample above, all the tests (throughout all the passes) were all negative for errors except the True Walking zeros (8-bit). This phenomenon has not been consistent. It does not occur throughout the passes consistently (i.e pass 1-3 has errors, pass 4 does not, pass 5 does, pass 6-9 does not etc). What's even more confusing is that, while typing this, i've run another test....and this is the first time i've seen MemtestG80 show 0 errors!!! OCCT also came up with 0 errors.

So, is this a software or hardware issue? What adds further confusion into the formula is that I tested my GTX 480 in my previous hardware (Gigabyte Mobo, 4GB RAM, Core 2 Quad Q6600, which I gave to a friend)....and it was behaving odd! It would either freeze while playing a game, or even freeze on desktop! .... this GTX 480 and prior equipment (with the same drivers) were working well together for years! My friend slaps his 6870 back in and no issues.

If you've read up till now, you can see why i'm really confused and wondering what the hell is going on. :)

Any feedback is appreciated.

r2rX :D
 
Last edited:
Thanks, cyberfish. At least it's running for now.....games which don't use too much vid mem seem to be alright. So i'll hold out until the GTX 680 is available. Any suggestions on what I should do with the card? Is it worthwhile attempting to get it repaired? Or is it a goner?

r2rX :D
 
Some people have had some success with thermal cycling (heating up to ~120C, wait a few minutes, then SLOWLY cool down). I have never been able to permanently fix a card that way, though. Usually they will start working but only work for a day or 2.

Other than that, I don't think there's much you can do, unless you have industrial reflow soldering equipment.
 
Back