Originally Posted by flipmatthew;11857493
Hi, my rig is an antec fusion 430 with a tx 650, 890gxm, phenom II x6 1055t (Stock) HIS HD 6850 (stock). Whenever i put the hd 6850 underload, after about 3 minutes, the screen will turn black, and the led will blink as if it is searching for a signal. i have to restart the machine. I've monitored temps, and it seems it happens around 70c (gpu.) I am also using a VGA monitor via dvi to vga dongle if that helps. i have upgraded to latest bios on my mobo and newest drivers. any suggestions?
I created this thread over at AMD forums:http://forums.amd.com/game/messageview.cfm?catid=279&threadid=142284
You're having the exact same problem as many others over there. Nobody is exactly 100% sure that it's a hardware problem, but it most likely is.
The solution we came to was to lower the GPU RAM from 1000Mhz to 900. Lowering the GPU to 600 instead also works (for me at least, but the performance penalty is bigger).
It might be the VRM's getting too hot, the voltage being too high (one user said he went with a pencil to lower the ram voltage from the stock 1.6 to 1.4 - but then again one user said he raised the GPU voltage from 1.15 to 1.2 and that that solved his problem, but these are isolated cases).
Anyway, just to be quick: to most people lowering the GPU memory to 900 solves the problem. To some though it doesn't. There is even one user with a factory overclocked card that needs to downclock the memory from the 1100 to 1000 to not get those hard crashes.
If it were a driver issue I guess this wouldn't happen. Especially not such a serious problem as a hard crash.
I can tell you I tried everything. Every driver version up to 10.12, using driver sweeper and CCleaner and TwlCleaner between installations; changing quality settings in CCC, editing the profiles files and putting Powerplay settings to all the same values so there wouldn't be any ups and downs in frequencies that could eventually be causing the problem; deleting and reinstalling the affected games, updating MB Bios, GPU Bios, newest Chipset drivers, etc, etc, etc. Even manually setting the fan at 100%, to no avail. It's not a GPU temperature anyway, it idled at a cool 36.º Celsius, and I ran Furmark for more than 10 minutes, at 69.º/70.º Celsius and could also run Unigine Heaven with Extreme Tesselation on, and 3D Mark 11 with no problems. Dirt 2 and Metro 2033 also without problems.
Crysis and Crysis Warhead and GTA 4: Episodes from Liberty City were not so fond. Some say it's because these games demand more bandwidth and the VRM's can't handle that load; a benchmark is just a scripted thing with less demand on the memory bandwidth part of the equation. But some people had problems with other games, which could mean their VRM's are even worse than mine.
It's been 19 days now since I answered AMD's questions and submitted the reports they asked me to, and I'm still waiting for an answer. All I got was a "your service request has been escalated to one of our experts who can better address your questions".
Of course lowering the ram to 900Mhz is not a solution since the card is most likely faulty; so I went back to the store and exchanged it for an Asus GTX 460. Some people have gotten bad luck and gotten two bad HD6850's in a row, so I'm not taking chances. I've lost too much time on this already, and AMD still hasn't answered my ticket with all the info they requested after 19 days.
Get another card; if you want to stick with AMD, go with a 6870, nobody on that thread came up with the same problem using one of those, or stick with the previous generation HD5850. Or go with Nvidia like I did.Edited by tpi2007 - 1/2/11 at 12:43pm