Jump to content
Sign in to follow this  
fabman_uk

AMD 6950 Overclocking

Recommended Posts

Okay so I have overclocked my computer, the CPU, GPU, RAM and motherboard. 3 seem to be working just fine, but 1 is not and that is the GPU. I have a Sapphire 6950 1Gb, its been running OC'd for a bit now and had been working fine however I recently developed a screen flicker while using my computer.

 

I had recently updated the drivers from 12.3 to 12.7 beta, I assumed this was the problem so I rolled back the drivers to 12.3 doing a clean install process (Uninstalling software full suite, deleting AMD folders, in safe mode deleting driver from device manager, booted into normal mode without internet connection, installed software and restarted). This did not fix it.

 

I took idle and load temperatures for my GPU while at the OC speeds, on idle it was 61 celcius and on load it was 80 celcius, which seemed a little high. I cleaned the computer the GPU and CPU of dust, checked airflow was good and that all cables and cards were seated correctly. This did not fix it.

 

I replaced the thermal paste that was on the GPU with Arctic Silver 5, using methylated spirits to clean the old thermal paste away. I now believe this has caused more problems as I also restored my default GPU speeds and voltages but my current idle temperature are now 60 celcius, which is dreadful considering it is not overclocked and using new thermal paste. I followed the methods recommended to me on the internet and by Arctic Silver for applying it, yet I guess it is wrong as it is not working. A rice grain shaped blob in the middle, I did not spread it but applied pressure with the heatsink and did not lift the heatsink off again once it was touching. This did not fix it.

 

Reading on the internet people experienced screen flickering with hardware acceleration on while using Internet Explorer 9, so I turned it to software rendering and it has improved dramatically, screen flickering only appears on boot up the computer and Sapphire Trixx starts. But when I run intense benchmarks (Unigine Heaven 3.0) for a while, my screen now goes blank and stops responding but it is still on. After a reboot I check the temperature log and temps are going to 82 celcius on my GPU and GPU-Z reports that GPU Temp. #3 goes over 100 celcius before the crash.

 

I can't tell if this is because of the original problem or because now I have replaced the thermal paste incorrectly it is causing an overheating problem. I have replaced the thermal paste a second time and things have not improved with the temperature. Surely 60 celcius is a high idle temperature for a GPU at default speeds, with new thermal paste and a cool operating computer.

 

Speedfan reports CPU idle temp at 38 celcius and MB idle at 30 celcius, with large CPU overclock and voltage increase. I can run Prime 95 on the heat generating option for hours with no problem. So I don't feel its a CPU, RAM or motherboard issue.

 

Summary of attempted fixes:

- Rolled back drivers from 12.7 beta to 12.3

- Cleaned the GPU and computer of dust

- Set GPU to default clock speeds and voltages

- Replaced the thermal paste on the GPU with Arctic Silver 5

- Turned off hardware acceleration on Internet Explorer 9

- Replaced the thermal paste on the GPU with Arctic Silver 5 for 2nd time

 

-Edit-

A quick test, it might be Sapphire Trixx. As a bit more testing and the problem doesn't seem to exist with it active, despite Trixx settings match the graphics card default settings. I am going to try to find out if I can overclock the GPU via a BIOS update.

 

-Edit 2-

Okay it seem that changing the BIOS clock speeds causes a BSOD on boot up. However in BIOS, activated 6970 shaders and using AMD overdrive set GPU to 840 and memory to 1325. Small increase but stable on quick test, I am not running Sapphire Trixx anymore. Temperature reached 75 celcius on load and on idle is now at 48 celcius... so Sapphire Trixx which worked for months became corrupt installing 12.7 beta? Now when used causes problems... strange. I will do further testing but looks like I solved my own problems. I will update as someone (maybe) will have a similar problem.

 

I guess I won't know if Arctic Silver improved temperatures, since Trixx inflated idle temps. Load temps are looking good so far, but will run benchmark for an hour plus. Hopefully doesn't rise above 80 and I know it worked.

 

-Edit 3-

I just enabled Sapphire Trixx and watched the temperatures rise about 1 celcius every 5 seconds, levelling out at around 59-60 celcius. I guess the software overrides power saving features and stresses it out to a high level, not allowing the GPU to slow by much and cool down even during slow parts of video games or just while surfing the internet. Uninstalling, now I need to find out how to OC my video card and increase its voltage, without stopping power saving features.

Edited by fabman_uk

Share this post


Link to post
Share on other sites

Okay so currently everything is running smoothly, however I am no longer using overclocking software like Sapphire Trixx or MSI Afterburner as both cause the problem.

 

I can unlock the 6970 shaders in the BIOS. Using AMD's OverDrive in CCC I can overclock the GPU clock by 5%, memeory clock by 6% and push the power control to +20% without any issue while benchmarking. I am going to keep these settings for a while to see if my graphics card is faulty or not.

 

What I am now trying to find out is if I can unlock AMD overdrives restrictions as I am at their max, or overclock my GPU by other means. AMD drivers will not work if the GPU has been overclocked in the BIOS, causing BSPD. The software Trixx and Afterburner causes problems for me when I use them, causing screen flickering and sytem crashing.

 

I know its a double post, but I think it needs to be seperate. Plus this thread is already at the top so its not as if I am bumping.

 

-edit-

I found a website that unlocked that gave me a registry entry that unlocked CCC OverDrive, and it worked after a reboot. I've set the settings higher and now I am going to test at 6970 settings. However I don't have a method of increasing voltage which will limit how much I can overclock.

 

Windows Registry Editor Version 5.00

  [HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Class\{4D36E968-E325-11CE-BFC1-08002BE10318}\0000]

  "PP_PhmSoftPowerPlayTable"=hex:6f,01,05,01,00,04,04,10,18,58,00,68,00,b8,00,e8,\
	03,e8,03,58,00,00,80,01,00,0f,00,00,02,08,3c,00,00,68,00,b8,00,00,18,01,09,\
	00,18,01,26,01,83,01,00,00,a0,0c,00,00,30,01,40,01,4b,01,00,00,00,00,00,00,\
	aa,00,00,00,78,00,00,00,60,ea,00,00,56,01,0a,00,00,00,14,00,00,00,00,00,00,\
	00,01,01,02,03,02,02,02,04,03,02,02,04,80,38,01,48,e8,01,01,ff,7e,04,00,00,\
	01,00,00,00,a8,61,00,98,3a,00,84,03,b6,03,00,00,01,00,00,00,50,c3,00,48,e8,\
	01,e8,03,7e,04,00,00,01,00,00,00,80,38,01,48,e8,01,01,ff,7e,04,00,00,01,00,\
	00,00,34,1b,01,48,e8,01,01,ff,7e,04,00,00,01,00,00,00,08,00,00,00,78,00,00,\
	00,00,00,00,00,00,00,00,00,00,00,00,00,00,00,00,00,05,00,00,00,78,00,00,00,\
	00,00,00,00,00,00,00,00,00,00,00,00,00,00,00,00,00,04,00,00,7a,00,00,00,00,\
	00,00,f0,d2,00,00,40,9c,00,00,00,00,00,00,00,00,00,00,00,7a,00,00,00,00,04,\
	00,70,11,01,00,c0,da,00,00,00,00,00,00,00,01,06,94,11,c4,22,54,24,60,09,3c,\
	0f,88,13,0a,00,d0,78,02,00,90,d0,03,00,03,40,9c,00,84,03,50,c3,00,e8,03,68,\
	3c,01,01,ff,02,98,3a,00,b6,03,48,e8,01,7e,04,02,98,3a,00,84,03,48,e8,01,e8,\
	03,04,84,03,11,00,00,00,e8,03,1a,00,00,00,4b,04,28,00,00,00,4c,04,28,00,00,\
	00
Edited by fabman_uk

Share this post


Link to post
Share on other sites

So your initial problem was with the GPU using a default BIOS but with overclocking? The first thing would generally have been to just run it compeltely stock rather than add in all the extra variables. It looks like the crashes may be caused by overheating - each temp monitor for the 6950 is a (generally quite reliable at high temps) Digital Thermal Sensor (DTS), on a different part of the GPU die. #3 is for the display I/O, so basically it 100 means a part of the chip is getting really hot - possibly due to not having any thermal paste over it.

 

As a first step I'd take off the HSF and see if the thermal paste did spread, redo the paste (perhaps do some spreading manually if you have to) and reseat.

 

Then take it step by step. Without any overclocking (in Catalyst or elsewhere) is the card stable using your current BIOS, and have the temps improved?

Share this post


Link to post
Share on other sites

Hey philo-sofa,

 

My first problem was screen flicking during general use not video games, I thought it was a GPU problem, so I replaced drivers, cleaned fan, replaced thermal paste and the flickering didn't go away. Turning off hardware rendering in browsers reduced the problem.

 

After some testing I had realised when Sapphire Trixx (OC software) is activated it causes screen flickers that never use to be there. I tried MSI Afterburner and it still occured. So I tried using AMDs own CCC OverDrive software, but that was limited in how much I could OC. I unlocked it with the registry hack, but it span the fan too slow for when it reach the high temperatures (around 34%).

 

After lots of testing it appears temperature is not the problem. I've had the card up to 90 and it didn't crash while it has crashed at 70. Also with no OC it idles in the low 40's so the thermal paste must be okay.

 

My conclusion so far and I think it will end here, is that I can no longer overclock my card to the same speeds as before. It is not the case of voltage as I cranked it up to 1.25 and it didn't improve stability at all.

 

So currently I am running my 6950 stable with the 12.7 beta drivers, the 6970 shaders unlocked, 880 GPU clock, 1325 memory clock, +10% board power, voltages are at default values and temperatures are reasonable now. When running at 99% GPU usage temps were around 75 celcius, which I can accept.

 

So I've taken a hit in performance (about 5-10% not sure) but gained back my stability and I can use the latest drivers. Not a terrible situation, I won't cry.

Share this post


Link to post
Share on other sites

I had the same problem with my HD5870 and the 12.1 drivers, where running multiple screens and MSI Afterburner to overclock the card would cause flickering of the secondary screen, and eventually turn the main display off - not crash the system.

 

Fixed by reverting the card to stock clocks in Afterburner, and it seems happy enough now.

Share this post


Link to post
Share on other sites

I had the same problem with my HD5870 and the 12.1 drivers, where running multiple screens and MSI Afterburner to overclock the card would cause flickering of the secondary screen, and eventually turn the main display off - not crash the system.

 

Fixed by reverting the card to stock clocks in Afterburner, and it seems happy enough now.

I am concerned that the card maybe getting damaged when it flickers like that.

Share this post


Link to post
Share on other sites

Possibly, it's hard to know for sure. I'd say that the likely component being damaged (if indeed any component is suffering) would be the memory, as when overclocking it's typically the memory that causes black screens and flickering.

 

That said, it's just conjecture from my own anecdotal experience - stock speeds shouldn't be a problem at all :)

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×