Well this weekend I decided to overclock my gtx 680 lightning. My settings are : 93mv core voltage, 100mv memory voltage, 50mv aux voltage, 300 power limit, 160mhz core clock offset, 520mhz memory clock mhz with a custom fan curve that never went above 70c except when I used Furmark it went to 75 then I started to see things on my screen and then the card downclocked everything down and it was like at 600mhz core even after I started another unigine run. So I closed everything, did a reboot and now its working fine but I have this feeling that I've somehow ruined my card. I played Starcraft 2 on the highest settings and everything was fine no artifacts or crashes but after that I decided to read some articles on cnet and the screen would become blurred for like 3-4 seconds then come back to normal. What do you guys suggest that I do to check if my card is 100% functional or not ? I am using the ln2 bios, 301.42 nvidia drivers.
I can't imagine that you "bricked" your card. If it got to hot it would either throttle (like it did), or shut down the PC to protect itself before sustaning damage. Also Windows Live never works right, if thats the only problem you have then I wouldn't worry about it to much.
This card is made for overclocking so I doubt that with a measly 100mv on core and memory it would degrade right? Temperatures are around the 70 mark in unigine maxed out except for 3d. Also as I have read my core clock readings are a straight line in 1 unigine run as well as the memory so I should be good right? The only time it oscillates is when the gpu usage goes down, core clock and memory go down as well, in rest they stay in a straight line for the entire time of the run. My score is around 57.4 with the clocks I posted above.
Why would anyone design a software with the sole purpose of increasing temperatures of cards to a point where they always artifact after 3-4 min?I'm talking about furmark. I uninstalled as soon as I saw what it was doing to my card.
Well I fixed the issues with windows live messenger it seems that when I disabled the FXAA in nvidia control panel the blur disappeared. Now if you guys are positive that furmark did not ruin my card in that 4-5 min run, then I'll sleep better tonight. The only reason I even used it was to find out my max temperature and design a custom fan profile using that information. Right now I have 3 points on my fan curve. One is for idle, 12% fan speed at 37c, one at 55% fan speed for 55c and last one at 100% fan speed for 82c. Is this fan curve alright or should I set up a better one? Max gpu temp in unigine is around 65. Memory and vrm in low 60.
A low GPU temp doesn't mean that your overclock is stable. You may want to consider the fact that you may have a less than average card when it comes to overclocking. Not every card can come out and do a 20% clock speed increase. There is also the possibility that your voltage adjustments are doing more harm than good for stability.
It wouldn't have ruined your card, in fact as far as i understand nvidia put a block into their drivers that prevents furmark from overloading the cards, at least this was the truth a time ago. If your card was damaged it most likely be crashing or artifacting even if you left it at its default clocks.
I would recommend you set your fan curve to hit 100 percent at 55c or above, when the hardware heats up it heats up incredibly fast, I would start earlier at 100 percent so it doesn't spike too high before the fan kicks in.
First I did my core clock overclock, increments of 20. At plus 201, I crashed within 5 seconds, so I backed down until 176 for the clock alone and it was stable. Then I did the memory increments of 100. At 600 I had artifacts so I backed to 570 it was stable. When I put them together they were not stable so I backed down even further to 160/520 and I saw no artifacts in avp benchmark, 3dmark11, unigine loop 10 min nor crashes. But for maximum stability, how long should I run unigine to make sure it's 100% stable? Are there any other benchmarks I could leave on running and check back later?
The new GTX6xx (Kepler) cards have a feature called GPU Boost. Your card will NOT allow you to go above a damaging temperature.
There are two main advantages to GPU Boost:
1) the temperature cap prevents damage, and
2) the card can increase the FREQUENCY of the GPU processing for more processing power (up to the temperature limit)
PREVIOUS to Kepler, it was possible to kill a card with Furmark. When graphics cards where made they simply designed them with cooling adequate for the normal games and programs that would run them. However, Furmark then stressed the cards beyond the intended use and they overheated and died.
I believe it's impossible to overheat a Kepler card due to the GPU Boost function. It will downclock as needed to prevent this (or overclock if not overheating for more performance.)