Overclocked CPU or GPU caused this crash?

[Dave3000]

I overclocked my MSI GTX 780 Ti to 980 MHz base clock / 1032 MHz boost clock using EVGA Precision. I then went to run 3Dmark 11 and just before the last test, before the loading, 3Dmark 11 went out of full screen mode to a small window mode and crashed on me. Yesterday I didn't have this issue in 3Dmark 11 at the same speeds. Nothing was showing inside the small window and about 30 seconds later that small window artifacted with a double image almost totally overlapping each other, even though the rest of the screen was fine. I could not reproduce that problem even with the base clock at 1020 MHz / 1072 MHz boost clock and I tested frequencies in between as I worked my way up to 1020 MHz base clock. The problem here is that the clock speeds are not static during benchmarking because of the turbo boost 2.0 feature on the GPU, so when 3Dmark crashed, was it possible that the turbo boost was operating at it's top speed at that time, while the other times it was stable the boost clock was lower?

My CPU is a 4930k and I have it set to an all-core turbo boost of 3.9GHz instead of the default single core of 3.9GHz. Prime 95 tested successfully for 9 hours and OCCT Linx with AVX and 25% memory size (I have 32GB RAM) tested fine for 1 hour. Also 4GHz with the same vcore on my CPU was no problem during 30 minutes of OCCT with the same settings.

So does it sound like my GPU was overclocked too far or my CPU can't handle an all-core frequency of 3.9 GHz without touching the voltage?

 

[Soulkeeper]

It could be anything. Your best bet is to isolate one overclock at a time.
I suggest you run the cpu overclocked with the vid card at stock first. Find it's 24/7 stable oc speeds first, then you can start pushing the vid card. These things can take time, so patience is a virtue

 

[Dave3000]

I ran the OCCT Linx test for 7 hours without problems. I checked the Event Viewer, and no WHEA errors during testing. With my GPU running at stock settings I did around 10-15 loops of 3Dmark 11, and that crash did not reoccur. This testing was done with my CPU running at 3.9 GHz, which is what it was running at when I had the issue with 3Dmark 11. Since I can't get this issue with 3Dmark to happen again even with my GPU overclocked, I'm not sure exactly what caused it.

 

[guskline]

How old is your PSU? Size of PSU? Cooling for cpu and gpu?

 

[Dave3000]

How old is your PSU? Size of PSU? Cooling for cpu and gpu?

The power supply is about a year old. It's a Cooler Master Silent Pro Gold 800w. For the CPU cooler, I'm use a Hyper 212 Evo and my 780 Ti is the one with the reference cooler. CPU's hottest core temp was 73c during 7 hours of OCCT Linx at 3.9GHz and auto voltage. GPU temps don't exceed 83c during benchmarking.

 

[lyssword]

I would bet its the gpu

 

[Dave3000]

Today I got the same problem again in 3Dmark 11, just after the Physics test where it says "Please Wait" while it's loading but it would not load the next test. Also I recently updated 3Dmark 11 because it said there was an update but then again I had this problem before the update. The first time it happened today my GPU was overclocked to 980 MHz base clock but it also happened when I went back to stock clocks on the card. So now I can rule out that it's an overclocking issue, unless I damaged my card from the overclock. Maybe it's a bug with 3Dmark 11? Also I checked Event Viewer and at around the time I start running the benchmark or sometime into it, Event Viewer says something about a Wudfrd driver failing to load, if that can have anything to do with the crash in 3Dmark 11.

 

[Dave3000]

I don't think the Wudfrd error message in Event viewer has nothing to do with the 3Dmark 11 crash because this time I successfully got through the benchmark without the crashing after the Physics test. Maybe I should just keep my hardware at stock settings from now on because overclocking makes troubleshooting take longer because first you have to find out if an issue with your PC is caused by overclocking when it might be a software issue. If I didn't have my system overclocked the first time I had that problem in 3Dmark 11, it would have saved so much time troubleshooting.

 

[Burpo]

It's possible the power delivery for the card dipped or was out of spec long enough to cause the anomaly..
Edit: Complaints about the listed Power Supply
The -12 V output failed to stay within proper range...
Doesn't CoolerMaster lab have an oscilloscope? Virtually every one of their units have issue with ripple and noise..
I've eyed out the times till 12V out of spec as about ~21ms for the Enermax, ~16ms for the Corsair, ~14ms for the Antec and only ~6ms for the CoolerMaster!

Could be you're right on the edge of what the power supply can deliver while overclocking the video card.

 

[Dave3000]

It happened at stock settings as well.

 

[aigomorla]

It happened at stock settings as well.

this is not a good sign....

This points out it could be a problem with the base hardware.

1. Bad install of 3dmark? (bad install of overlaying drivers... bad software install of some sort.... did u install your drivers while overclocked?)
2. Ram (most of the time i realize its a ram related issue.)
3. Bad allignment on cpu with pins... (yes sometimes this can cause ram issues)
4. Bad voltage on board settings at stock..
5. Bad board.
6. Hickup on PSU.... maybe a bad rail...
7. Temperature related problem on either cpu or gpu...


Its going to be a long long debug from here... and without spare parts, its going to be a hassle..

 

[crashtech]

It might be best to try and eliminate a driver issue first. Might want to try rolling back your system to a time well before this happened. Documenting which driver versions are in use right now vs what is being used in the rolled back OS may help the troubleshooting process.

If the problem is hardware related, troubleshooting it may be difficult if not impossible without a generous supply of spare parts.

 

[Dave3000]

Well I reinstalled Windows 8.1 about a week ago. The only thins I did different this time was instead of installing the Intel Rapid Storage Technology driver software, I went to the device manager and just updated from the MS SATA driver to the IRST driver, and I didn't uninstall the MS SATA driver. Also I installed the latest Intel Management Engine driver for my motherboard instead of the previous one. I also have Classic Shell installed but I don't think this could cause the issue. I also flashed my MSI GTX 780 Ti's bios to the EVGA's GTX 780 Ti's SC bios since they both use the reference design and I wanted to run my card's base clock at EVGA's SC speeds (980 MHz) for there reference designed one, as MSI Afterburner and EVGA Precision only let you overclock the core boost clock, not the base clock. After flashing back to the original BIOS for my video card, the problem started happening. Either the overclock damaged my card or something is missing or didn't get configured properly during the flash process back to the original bios for my card?

 

[crashtech]

Do you have a spare gpu?

 

[Dave3000]

Well I have an AMD R6450 lying around. But it uses different drivers and it's not the same video card as my 780 Ti, so even If I don't have issues with my R6450, that could mean a damaged video card or a video driver issue. I would have to keep running 3Dmark 11 very frequently to find out the problem since it's an intermittent issue but it happens at the same spot, just after completing the Physics test where it says "Please Wait". I don't know if it's my SSD either because one time it was taking forever to transfer a files to it from my 1TB HDD, around 16-32 KB/s at almost the end of the transfer and I decided to cancel the transfer because it was taking way longer than it should have. I thought that using IRST driver fixed that issue because I didn't experience it again after installing the IRST driver, but now I'm not so sure if this is issue related.

 

[Dave3000]

I want to add that when I run the Valley benchmark, I get white flashing artifacts in the distance where the mountains are, but this issue didn't not happened with a very old video driver, but does with the video drivers after that one.

 

[Dave3000]

I just tried my AMD HD6450 card and the problem did not reoccur in 3Dmark 11 but I only ran it twice. It takes like 20 minutes on this card to run 3Dmark 11, much longer than with my 780 Ti. I only ran it twice so far on the HD6450 so that does not really tell me anything yet. I can run 3Dmark 11 for 20 runs and the problem might not occur with my 780 Ti, but can occur if I run it frequently enough. I don't want to be looping 3Dmark for a full day or two since this card is much slower and takes like 4 times longer to complete 3Dmark 11 than the 780 Ti.

 

[crashtech]

Would I be wrong to assert that this started happening after you returned your 780ti to the stock BIOS?

 

[Dave3000]

Would I be wrong to assert that this started happening after you returned your 780ti to the stock BIOS?

No, you are correct. However I have been running 3Dmark 11 more often than usual as well than before flashing my card's bios so I can't blame the card yet. I suspect either my motherboard, SSD, the way I installed the IRST SATA driver, or my video card. I should not have flashed my video card's bios with another manufacturer's bios even though that bios came from a card that uses the same reference design and cooler. My CPU and RAM is stable during hours of stress testing with Memtest 86+, OCCT Linx, and Prime 95 so I can't put the blame on the CPU and RAM as primary suspects.

 

[crashtech]

Oh, I guess you did say that this started happening when you did BIOS flashing.

After flashing back to the original BIOS for my video card, the problem started happening.

Maybe, just maybe, something went very slightly wrong during the process of returning your GPU to its stock configuration. Since you successfully flashed it to the EVGA BIOS once before, it seems to me that the risk for you to do it again would be minimal. Not something I would suggest to just anyone. But if it fixes the problem, you know you need to concentrate your efforts on properly flashing the BIOS back to stock, if that's really what you want.

 

[Dave3000]

Update: I reinstalled my MSI GTX 780 Ti and I reflashed the bios with the stock bios for my card using nvflash but this time without any extensions to the command. That didn't fix the issue. I was looping between the Physics test and the Combined test and about 10 minutes later the problem showed up again. It showed a little square window on the upper left of my screen just showing black with the title 3Dmark 11 Workload and doing nothing. My PC did not crash though as I was able to end 3Dmark through the task manager. This was after reinstalling Windows 8.1 without installing additional software, just the drivers except the IRST driver. I can rule out the way I installed the IRST driver on my previous Windows 8.1 installation the cause. If you look between tests in 3Dmark 11 there is a quick flash of a small square window on the upper left of the screen, that's the window I'm talking about.

 

[crashtech]

My suggestion was to actually flash your card to a point where this problem was not happening, which sounds like it was the EVGA BIOS, oddly enough. Perhaps not a permanent solution, but it might reveal a problem with the stock BIOS you are using, or the process used to restore it.

 

[Dave3000]

My suggestion was to actually flash your card to a point where this problem was not happening, which sounds like it was the EVGA BIOS, oddly enough. Perhaps not a permanent solution, but it might reveal a problem with the stock BIOS you are using, or the process used to restore it.

It does it with the EVGA bios as well. It even does it with a newer bios version for my video card. It does seem to happen more often after the Futuremark SystemInfo 4.30 update, as I can get the issue to reappear more often after the 4.30 update. It can take anywhere from 10 minutes to a few hours to reappear and it's not during an actual test, but right after finishing a test before loading the next test.

 

[crashtech]

We might assume that something is wrong with the GPU or its firmware. Do you have another PC in which it could be tested?

 

[Dave3000]

I was thinking about switching to an i7-4790k and a Z97 motherboard. At least I rather do that than buy a new X79 motherboard. I suspect that the motherboard could be the problem as well. I doubt it's the CPU since it passes stress testing for many hours even overclocked to 3.9GHz for all-core loads as by default it's only does 3.9GHz for single-core loads.