Overclocking Help/3DMark Vantage Issues

[montereywill]

Not sure if this belongs in the CPU or the GPU Section. Anyways here are my specs first:

Windows 7 Ultimate 64-bit
XFX 790i Ultra Motherboard
Intel QX9650 @ 4.00 GHZ [1.35v, 400 FSB, x4 SPP/MCP, 48C Idel Temp]
(Koolance CPU-360 CPU Block)
OCZ SLI-Profile 1800 MHZ memory at 8/8/8/28 @1.95v [unlinked]
BFG Tech GTX 280 x2 SLI [stock voltages, 725 Core, 1464 Shaders, 2650 Memory, 34C Idle Temps]
(DangerDen Tieton GPU Waterblock x2)
Single Loop with Swiftech MCR320 Radiator and Koolance RP-1000 pump/reservoir


Now for my problem. What I am trying to figure out is whether or not I fracked up my Video Card(s). I have been slowly creeping the video card settings upwards from the stock settings of 602 Core, 1296 Shader, 2214 Memory. I am using the EVGA Precision tool, and I have been monitoring temps. CPU temps while running either 3DMark Vantage, PCMark Vantage, Everest Ultimate or the Intel BurnIn Test have never exceed 72C on the CPU or 60C on the video cards. I have had my share of system reboots while I was working on the CPU Overclock, but I ran Everest and the Intel tests for stability before jumping on to overclock the video cards.

My problem appeared last night. I had been running multiple consecutive benchmarks of 3DMark Vantage testing the clock settings 725 Core, 1464 Shader, and 2650 Memory with no issues. Then I used EVGA to raise the clocks some more, and once I got my clocks set at 750 Core, 1500 Shader, and 2700 Memory I started 3DMark again, and while it was running the first test the PC reset. After it rebooted, I dropped the GPU Settings back down to previous values and started 3DMark again, and that was when I started getting the issues immediately. The test starts and while the lady is on the jetski, the water just looks like a bunch of rendered peaks straight up out of the water line and you cannot see the jetski. Other times I will get an error from 3DMark that says something along the lines of the program couldn't be rendered. So I decided to use the Crysis test. Crysis starts fine, and even runs with no visual issues at Very High settings.

What do you all think my problems are? And yes I still have these problems even if I set all clocks back to stock, which is why I am wondering if I killed a card, or did 3DMark somehow get corrupted due to the system crash during overclocking. Please advise.

 

[NumericalMethods]

The symptom of the water on 3D mark vantage you described happened to me when the RAM on one of my GTX260s let go. I'm not saying that you have the same problem but it is definitely time to test for it.

Test each card separately using something like the OCCT gpu test. If you have a fault with a card it will start throwing up errors fairly quickly. You may want to check your drivers are installed correctly again if you made any major changes to your system before the error occurred. Finally, check your SLI bridge for secure installation. I've had that cause problems with my SLI setup too!

That's the major things that I can think of checking. Good luck.

 

[montereywill]

Without unplugging one card from the system I ran OCCT in both SLI and Non-SLI mode. I assume that under SLI both cards have their respective memories tested by OCCT, where as under normal mode, only the first slot gets tested. Anyways, I think my results are still conclusive as under SLI-Mode OCCT said I had about 60,000 +/- memory errors after 10 passes, whereas under normal non-SLI mode OCCT reported 0 errors after ten passes. My assumption is that the second card has failed in one manner or another. Regardless of the settings on the card (stock or otherwise) I get the odd issues in 3DMark and with OCCT I get errors. I even swapped out my SLI bridge to make sure that was not the issue. Time to buy something new it seems.

 

[NumericalMethods]

Boo

The conclusion you have come to seems reasonable. Bad luck, your overclocking was going very nicely up to that point, I'd say. Your experience is precisely the reason that I'm less enthusiastic about GPU overclocking compared to CPU overclocking. It's amazing that you can seemingly kill stuff by just raising the clocks. Hope you get something sorted out.

 

[montereywill]

To be quite honest I was unaware that simply raising clocks on a card would kill it (or therein some part of it). I was under the impression that it was a lot like overclocking a CPU. You set the settings, then try to run something with that setup. It runs or it doesn't. Obviously if you try to run at a particular setting without sufficient cooling that would kill the hardware, but simply playing with numbers does not seem like the likely culprit. Anyways, new cards arrived today, and as soon as my new waterblocks arrive (supposedly on Saturday) I will be back to playing with the graphics, but clearly not as aggressive this time. Wish me luck!

 

[montereywill]

By the way...for anyone who's interested, I got one of my BFG Tech 280 video cards clocked to 760 Mhz Core Clock, 1535 Mhz Shader Clock, and 1330 Mhz Memory Clock on a water cooled setup. I am running a single loop, and my GPU temperatures idles at 43 degrees celsius and rises above 55 degrees Celsius, while my overclocked Intel Core 2 Quad QX9650 running at 10x400Mhz (4.00 Ghz) at 1.375 vcore idles around 45 degrees Celsius and the highest I have seen it go underload running Prime 95 is 61 degrees Celsius. Too bad my second BFG Tech 280 GTX crapped out somewhere around 725 Mhz Core Clock, 1464 Mhz Shader Clock, and 1300 Mhz Memory Clock.

 

[NumericalMethods]

To be quite honest I was unaware that simply raising clocks on a card would kill it (or therein some part of it). I was under the impression that it was a lot like overclocking a CPU. You set the settings, then try to run something with that setup. It runs or it doesn't.

snip

Meh, I know exactly what you're saying. RAM always seems more fragile than CPU's for whatever reason. Maybe the RAM on the card was marginal and pushing the clocks was enough to finish it anyway. Glad you got a replacement sorted out so soon