6970M Undervolting

Yeah memory 1060 gives me this cut in half screen, and 820 is just 5 minutes stable then the Game just crashes, same with Crysis 2 but there in about 30 minutes, tried nearly every drivers but doesn't matter that much not every card is the same maybe an OS reinstall COULD help

 

Hey guys, I can't find the "log file" after I benchmarked.

Do you know where the log file is located?..

 

Hello guys, i've been using my AMD 6970M in my Sager with the undervolted profiles for around two months and they really take the temperature down to 73C while gaming (FNV) @680/900. Recently i noticed that my GPU 'locked' @100/150 and i'm unable to reset it to stock values. I tried reinstalling drivers and the ATT ,but all that without any effect. Any ideas how to fix this? Thanks in advance!

 

Make another profile with the stock settings,

680/900 @ etc voltage, and save it to a profile.

Assign hot keys to the 100/150 setting and the 680/900 setting,

and you can manually switch around the two settings, and it works so great for me now.

These cards are hardware-wise locked to not go less than 0.750V, so

100/150 @ 0.750V

and my cards take up to 0.905V @ 680/900 without crashing.

Try to find your 3D settings with the lowest possible voltage.

 

It it working fine with both of the profiles i have. I have two profiles for 100/[email protected] and 680/[email protected]. The point is ,that I cannot stock clock it neither with the hotkey, nor the right click menu -> Hardware -> 3D Profile (Gaming Profile). And the Auto Overclock doesn't work too. It just stays there...

 

The auto-overclock is kinda funky for me too.

Hmm... I don't know whats going on, but it seems like I won't be the one able to help you..

as I've only used this tool for a day or two..

 

have u flashed bios with that voltages and it doesnt work??it remains stock voltages??

 

Hey, Does the Auto Overclock depending on power status work for you guys?

I always get a BSoD when I enable it. ATIkmdag.sys always crashes for some reason when I change power status to battery mode.

Using a 6990m

 

Nope =[. Flashing the vbios had no results as it just took the stock voltages.

 

Do you had have Overdrive enabled? If yes, then you have to reset the settings in the overdrive panel (CCC).

 

So far finding:

1v - 735/900 max
1.010V - 790/910 max

then I had to bump the voltage all the way to 1.060 to net me extra OC headroom so far:

1.060V - 800/950 is very stable. Can and will go further on the core soon.


I am thinking the voltage on this card does not go up or down in small increments but in much larger steps! Hence voltages between 1.010 - 10.50V made no difference to stability at anything higher than 790/910.

 

Nice feedback, thanks I gonna try 790/900 at 1,010v too, with my 6990M.
Btw have you tried 740/900 at 0,96v? It's fully stable with my 6990M! Hundreds of hours playing and no problems whatsoever.

Best Regards to all

 

Nope but I tested 810/950 at 1.060V and I had black screen after 10 mins of play with dx11 enabled all settings max. Frames were averaging 25-30! Still your clock at 0.96V seems to prove my theory. I bet the actual voltage used at 0.96V is the same if you set 1V. It is only after using 1.010V-1.050V that the next incremental increase in voltage is triggered. Also I can run stock clocks 680/900 at 0.91V which is probably where the increment before 0.96V starts.

Then I backed the memory down to 900 and tested 830/900 at same voltage and frames were the same and stable! In my opinion the memory therefore doesn't need to be OCed. Firstly is doesn't net enough of an increase in frames to be worth it and secondly it sucks up lots of power.

Increasing the core is much more effective and stable, cooler and gives better performance. My max sensor after one hour playing was at 87C coolest at 78C.

So basically even at such a high OC at 20mhz bump in core clock did what a it took at 50mhz bump in memory!

I can imagine you will have similiar results with your 6990M although your card has more firepower than mine so might put the bandwidth under a little more pressure. I must be running close to your performance at 830mhz aiming for 850mhz today if I have time at same voltage.

So it seems voltage increments happen every 0.050V or so. eg 0.91V, 0.96V, 1.010V, 1.060V and then 1.1V(a little odd perhaps) maybe 1.1V is no better than 1.06V only further testing will tell. I will of course eventually try 1.110V just to make sure....

 

I agree that the Memory doesn't need to be OCed. I tried 790/900 at 1,010v, and indeed it was very stable I also have the same opinion, regarding the voltage increments, for example: 715/900 at 0,955v was unstable, but 740/900 at 0,96v is fully stable.

I will try now 680/900 at 0,91v, like you suggested (I had 0,92v and was stable).

Keep posting your results and tests,
Best Regards to all

 

well today I had hard crash. Now computer dies mid boot. Works in safe mode. I think my graphics driver is corrupted....

 

Dont worry, download drive sweeper and clean all related to AMD ATI. Then install, 11.11c, 11.12, 12.1 preview or 11.12 beta 18Nov (best until now).

Best Regards.

 

Mine is running at 700/900 at 1.00v.

Max temp 79'C. (After 2hrs of BF3)

I'm living in Singapore, so the ambient temp will be higher.
I am not using any cooler either, just jacked up the back for more airflow.

So what you guys think of my temps?

 

hmm I think laptop is dead or at least partially dead! I did driver sweeper but even with no driver the machine still dies at same point mid boot. Even with no OS and booting from windows dvd it also dies at exact same point!!

 

Damn...Any reason why that happened? Did you tried a linux live CD (any distribution..., example: Ubuntu/Fedora)? If it works, try installing it normally, in order to install/test the use of your GPU.

Good Luck,
Best Regards to all

 

I think pushing the core at 850mhz did it! The poor thing couldn't take it

I am pretty sure it is hardware based problem as the shutdown occurs when windows tries to load the drivers I assume. Which means probably either sound or graphics. Due to when it died I am going to assume it is the graphics.

Thinking of getting P150HM with cheapest processor and with 6990M chip. Hopefully can configure without ram,HDD and ODD

 

I bought mine without HDD and ODD, but I'm not sure you can purchase without RAM, but since you already have 8Gb 1333Mhz, buy +8Gb to fill your new P150HM with 16Gb Just a suggestion

Best Regards to all

 

See below EDIT.

I can report also that the 6990 works with this undervolting. I have mine at 715/[email protected] and I have temperatures on my 6990 down by at least 5~6C but more importantly it's more stable whilst gaming.

I ran my metro2033 and crysis and the fps are nice and high and smooth. Whilst, at 1.10V at 715/900 I can report more heat most definitely but the fps are not smooth...

EDIT: Can report after long gaming session it does have issues.
So have returned to 1.10V at 715/900 and it's much more stable...
Guess the 6990 is not quite the same as the 6970.

EDIT 2:
Noticed that the fps is worst with the undervolting so left it back at default. Tried it twice now.

 

how do you undervolt this card? i changed the voltages in rbe and flashed but when checking in cpuid hw monitor it was still showing the stock voltages.

 

Best to use Ati Tray Tools. There you can change voltage and clock speeds. Also you can manually input new voltages to use.

 

I finally got the voltage change to work via rbe, rebooted and ran furmark and the voltage went from 0.9 > 0.91.

gonna see if it black screens.

I'm going to return the 180w psu i bought and get some new thermal paste and pads.

 

Hi,

I'm running a Clevo 6970 in my G73 now.
Works like a charm.
I've tried ATi Tray tools to undervolt. Earlier I always modified the vBIOS of the 5870.
The thing is, I see now differences at idle clocks below 0.9V. Checked 0.85 - 0.8 - 0.75 V and other combinations. For the 3D clocks it's seems to work. Was able to run 680/[email protected] without errors. At 0.9V the driver resets. 0.96 and 1V are the same for me. With 0.95V the temps drops about 2-3°C and lower fan speed.
Here are my test results.

Btw.: I modified the idle voltage in vBIOS to 0.8V and 3D to 1V. But the tools still shows the stock 0.9V.
Switching to 3D mode changes to 1V correctly.

 

That's one insane undervolt, congratulations!
I am really amazed, wow, I wish my retailers sold the 6970M or 485M..

 

I got my 6970 down to 0.695v and lowered idle speed temps down another 3 degrees.

 

IMO you could set it to 0.001V without effect.
There is still a lowest value the VRM accepts.
At 5870m it was like 0.8/0.85.
I see no diffs at my 6970m with values below 0.9V.

 

I finally found a reason to undervolt my 6970m: I discovered that it hit 98°C while playing Skyrim (without FN+1 btw). When using FN+1 while playing Skyrim for half an hour I got 87°C.

So, I started to do some undervolting using ATI Tray Tools and using FurMark to test the stability and temperature.

When I got as low as 0,95 V the video driver got reset during the benchmark. So I had to put it on 0,96 V. I used stock clock settings: 680 core/900 mem.

Now I was ready to do a 15 minute burn in test:
At the 5 minute mark the GPU temp just hit 98°C which I think is too high for prolonged periods of time, so that's when I used FN+1 to cool things off a little.
After 4 to 5 minutes the temperature dropped to 89~90°C and stayed there until the end of the test.
The video card driver didn't reset during the test so I think it's stable under heavy load.

While gaming the temps seem a little higher since I just got the P170HM, but it's still idling at 43°C.

I still have to test the undervolt settings in games if there's any improvement.

 

Good call on the undervolting, that does seem way too high.

I undervolt to 0.95V @ 680/900 and do not see temps above 76 while gaming

 

Yay! Undervolting to 0.96V (see my previous post) did the trick in Skyrim.

Previously I got 98°C, but now it hovers mostly around 85°C when a lot of stuff is displayed in caves (spider webs, moss, torches/fire, etc.) and around 82°C in open places. I think I played for 1½ to 2 hours straight.

This is without using FN+1 btw.

 

I would try to apply new thermal paste as well. Those temperatures are very high. You should get considerably better temps after a re-paste

 

Anyone tried flashing the values to vbios ; o?

If it works, I'll re-flash my amd 6970m..

 

Nvm.. tried it again and re-applying Overdrive and then disabling it seems to work ; )

Maxed 90*C on Furmark at 680/900 @ 0.91v

Bios Clock modifying only seems to work for the max clock.

Modified Bios settings :
680/900 @ 0.91v
500/900 @ 0.90v
400/900 @ 0.85v
300/900 @ 0.85v
100/150 @ 0.75v

--

Only 680/900 seems to work though...