I'm currently trying to overclock this http://www.newegg.com/Product/Product.aspx?Item=N82E16814127737. I know that overclocking is generally a hit or miss depending on how heavily binned the specific GPU you get is, but pretty much everyone who attempted to get an overclock on a 7950 has said they have gotten over one gigahertz on the core with stock voltages and even more headroom thereafter. Whenever I run OCCT's GPU stress test on minimal overclocks, my computer locks up (drivers don't crash or anything) after 15 minutes or so and I have to do a hard reset to get back into Windows, but it runs fine with stock clocks. Does this mean that my card is just past the line where it's binned well enough for stock clocks, but won't work well with any type of overclock, or is it something differently entirely? Is stress testing nowadays still completely neccesary, or is testing overclocks on demanding games more realistic?
Why are you using OCCT to test the stability of your graphics card? Progams like that one and Furmark/Kombustor Overload the GPU and are not indicative of a real world basis for GPU stability.
Get yourself Unigine Heaven 4.0 and Valley 1.0 and set them up on the maximum settings with your screen resolution. Run Heaven first for 30 minutes and monitor GPU temperatures with either MSI Afterburner or GPU-Z. Preferably MSI afterburner, as you can put an OSD on top of Heaven and Valley while they are running. Then run Valley for 30 minutes doing the same as Heaven. If your screen doesn't freeze and you don't get artifacts while running either of those in a loop with good temps, your overclock is stable for normal game use.
This is the testing method I used to get my Asus HD 7870 DCII v2 to 1200 MHz core and 5,500 MHz VRAM stable at 1.250v. 1200 MHz is slightly higher than average for most HD 7870's, most can only get to 1150-1175 MHz stable with even fewer able to reach 1300 MHz or higher.
Thank you. I thought OCCT/Furmark/Kombuster were the ways that most people completely confirmed their overclocks and considered them "stable", but now that I'm doing more research on stress tests, it seems as though they're unneccessarily demanding. I'll use Heaven and Valley from now on.
The key is they are actually 3D rendered objects within a game engine so they are more of an accurate representation of actual gameplay as well as GPU temperatures over a period of time. GPU temperatures are very important if you are playing a game for 2-3 hours or longer at one time at your max overclock.
Using your method, I've gotten a stable overclock of 1075 on the core and 1575 on the memory at 1.25 volts, but I'm a tad concerned about my temperatures. In Heaven and Valley, I peaked at 93 celsius, but on a gaming load, I peaked at 89 celsius and averaged around 82. My specific card has a single, overclocked bios which boost the core's votlage up to 1.25 regardless of what I set it to in MSI Afterburner, so I can't seem to undervolt the card. Are my thermals fine or should I be concerned about the card in the long run? For case airflow, I have a fan blowing directly on the card from the side window and more fans generating generally positive airflow within my case, so I don't know what else I can do to help the card run cooler.
If you download MSI Afterburner v22.214.171.1246 Beta 12 you can configure MSI to turn off power play and create your own 2D and 3D clock rates that will power down the card when just browsing or what not, but go to full overclock when you load a GPU accellerated program, like a game.
89*C is definintely too high, you'll want low 70s at the most with the custom fan speed profile kicked up all the way to 80-100%
Thanks. I'll try to undervolt the card until I get an aftermarket cooler of some sort for it.