[Guide] Installing AMD Radeon Mobility 6970 / 6990 in your M17x R2 (Single GPU and CrossfireX)

Cool, so I did a format and created an image with all drivers installed minus the video card.

Flashed my GPU's to stock voltage and still no luck with the 11.11 drivers (lucky I created that image eh? )

Using 11.10 + 11.11 CAP1. Still to try a game....

Got a bit of a boost in 3dmark11

 

ati profiles 11.11 cap2 is out. i am using this now with 11.10 drivers. its obvious ati removed 11.11 GPU drivers only. 11.11 CCC can be downloaded seperately.

 

I am experiencing a strange issue.

My primary GPU does not appear to upping its clocks to 3D performance, but the second GPU does....

Tried reinstalling drivers etc, but something appears amiss. Do you think Overdrive may be causing it? I inadvertently enabled it to have a look and I am thinking this is when it started happening. I have disabled overdrive.

I might re-flash the card.

Any ideas? anyone seen this before?

EDIT:: Well, its seems that it is the quality software released by AMD.....I upped the GPU clock by 5 Mhz using Overdrive and lo and behold, it magically started activating 3D clocks..

 

Is there a way that you can specify which card is the primary and which is the secondary?

Its weird, when running a game in crossfire, my secondary card has higher usage then the primary card.

Almost as though my secondary card is acting as the primary?

 

Douse - this may sound like strange advice, but swap the cards into opposite PCIe slots if you have not already tried doing so. I had a similar issue with the M18x and 6990M cards and for some reason it made the problem go away even though re-seating them had no effect.

With respect to upping the clocks and having the problem go away, I had a pair of defective 6970M cards in my first M18x. They both behaved in a similar manner to what you describe with the one card. It took a month to get replacement cards sent to me because they were on backorder. While waiting for them, I did some experimenting and found that using RBE to set the clock speeds for all power states to the maximum defaults clock speeds (thereby defeating AMD PowerPlay) the cards performed flawlessly. If swapping slots does not fix the issue, using RBE to make these changes should resolve it for you. I have found that PowerPlay technology sometimes does not work as advertised.

 

Cheers for the advice Mr. Fox

They have been behaving very strangely since the format so I am fairly certain it is not hardware.

After a lot of tinkering I got the card activating 3D clocks at stock clocks.

Problem is I then had very poor crossfire performance in games, like 30% GPU usage.

So after I had enough I rolled back to the vanilla image and installed 11.9 drivers and caps and it seems to be functioning correctly now. 80-90% utilisation in skyrim.

I dont know what it is, but 11.10 and 11.11 seems pretty flaky..

 

Yes, I found 11.10 to be very flaky as well. Display brightness functionality was broken on my M18x with 11.10 and I had frequent driver errors reported in Windows Event Viewer. I found 11.9 much better. AMD really needs to get their act together and start producing better drivers. They have become very inconsistent based on my experience.

 

Yeah I agree. I am pretty shocked at just the amount of problems I have had.

I am thinking that when I upgraded from 11.9 to 11.10 pre-format it must have been using some 11.9 attributes as 11.10 now does not work in a fresh install....where it was working before.

Also I am not touching AMD overdrive with a 40 foot pole. If I want to overclock I would rather flash the cards.

 

That is the best approach for a permanent solution. I also had excellent results with Trixx for on-demand GPU overclocking.

I PM'd you with another recommendation.

 

Mr.Fox, if you are running 11.9, what caps do you use? Have you stuck with 11.9 cap4?

 

I think I may need to jump back to 11.9. I get so many driver crashes in 11.10

 

@Douse - Yes, 11.9 CAP4. That is what I'm using on the R2.

@DR650SE - Probably a good idea. I found 11.10 to be very inconsistent. Sometimes it worked great... other times simply terrible results. It was hard to gauge since my 6990M cards were defective (and eventually died), but now I am seeing enough feedback from others that it seems apparent that my issue with video cards was separate from the driver problems.

 

Found that 11.11 CAP 2 is really good for BF3.

GPU utilisation >95% on both cards most of the time.

 

it also fixed microstuttering on saint row the third during crossfire as they promised. initially i though it was a power issue till i found and installed the new profiles.

 

Out of curiosity, how hot are other peoples 6990's running?

After a good gaming session of BF3 (~4 hours) my Shader can hit around 85 degrees.

It doesn't bother me as I know it is within limits, I am just curious.

This is at stock voltage. I have under clocked CPU to stop throttling in BF3.

I am about to drop voltage to 0.96 now that it is running OK again.

EDIT:: Another question, anyone know of a fix to get HWINFO32 to detect the second card without have to muck around with GPU-Z?

 

i havent found a way on my 6970s. but if i remember well the developer removed a feature from hwinfo from a version and afterwards that was raising the gpu temps by its own. this feature was constantly detecting the second gpu but was raising temps so we asked him to remove this. Anyway you can ask him in his thread in NBR

 

disable ULPS and the card will show up.

edit: by registry or by installing saphire trixx wich has an option for disabling it.

 

yes thats an option. but does it worth spending more power for a few minutes of temps detecting?

 

Thanks CptX, that appears to have worked! Thanks

Aristo, you are right, normally it wouldn't really be an issue but because I rely on HWINFO32 to read temps and adjust fan control it is kind of important for me

 

Yeah, those 6990 ES cards require manual fan control. That is why I got rid of mine. Douse, do you have problems with shut down and sleep?

Also, do you have to install each CAP in order, or does the newest one have all of the older ones included in it? For instance, does CAP 2 for 11.11 have all of the contents of all of the CAPs that came before it?

 

Hey Turbod, yeah I do need HWINFO to control the fans, but it does the job really well and I have it all set up to start with windows.

I have no problems shutting down and restarting. I have not tried sleep, as I don't use it.

I did have some problems with the monitor turning back on after it turned off to save power, but it barely occurred and I think that was driver related. I have just set it to dim rather than turn completely turn off but I haven't tested it again since formatting.

Audio over HDMI works like a champ as well!

Overall very happy with the cards, but I think they are running a little hot so I might do a repaste.
I just got a whole sheet of Fujiypoly 0.5 pad so I can be a little more liberal and cover the whole memory chips as I barely had enough to last time

 

I have no audio over HDMI with this card, sadly. That is the only thing that doesn't work like it should. The FujiPoly is really good stuff. Dell's crap is really spongy, while the FujiPoly is very dense.

Are you only putting pads/paste on the die and RAM? There are several other components that I think deserve attention. If you don't have any 1mm pads, you could just double up two .5mm.

Here is what I have done to my 6990, and it runs very cool, no matter what clocks I run it at. It would be very easy for you to do this stuff (unless you have already).

6990M Thermal

 

Hey TurbodTalon,

The latest CAP profile will have the past cap profiles inclusive, so all you need to install is the very latest.

CAP 11.2 is supposed to be quite good for BF3. I have seen some pretty good results on my end as well as others post pretty good results on the forum too.

Cheers.

 

If it also needed in case of overclock of the cards using third party oc utilities. Not required if you use the embended OD in CCC (limit rised via RBE mod and vbios flash).

I need to have Ulps disabled becouse i rum my cards @0.96V. If i want to OC'em i rise the voltage with trixx and OC with Catalyst OD. This happends rarely since the cards doesn't need to be OC at all, just for benchies

 

Turbo, what temps do you get with pads and paste on the other stuff? maxxed out yesterday around 80*c playing BF3 all day. Cards at stock clocks and 1.0v

 

I've been playing Skyrim and Shift 2 all day long today, and I only hit 77°C. That is at 1.1V and 800/1000 clocks. Not sure about my ambient temperature here in the living room, but I have been hot all day.

FurMark is probably a better benchmark though, since we're playing different games. I might do a 10 minute FurMark run and see if my stuff is even really all that it is cracked up to be.

 

I might tomorrow as well. See where Im at. But since games are the primary purpose I think we're sitting pretty. And they are cooler than 5870s

 

Here is this.

Imageshack - 76199245.png

 

Hmm. The Furmark stability yesterday i did topped out around 85 but when i came back it had closed out. Its was in the background. I was running it as full screen to use with cards. I won't bother much with it because it involves swapping in my benching ssd, but temps in game are great and temps in benching were good when i was benching.

All in all i love there cards.

 

If you're topping out at 85°, it isn't worth messing with. You are running lower voltage, and two cards in CrossFire will never get as hot as a single GPU. Just like with everything else we do, I guess this one will benefit some a lot more than others. If it ain't broke, don't fix it! But I do wonder if you pasted and padded the other components how much lower your temps would be. Maybe someday if you're bored.

 

Yea, I might, but for now I'm in no hurry. Now it's more time for me to enjoy the system since I've spent so much time benching it for all it's worth. But always trying to lower temps. But your right, crossfire will always run cooler than a single card with 100% of the load on it.

 

It's about time you started enjoying it buddy. I think your R2 will love you even more than it does now, LOL. That poor little guy has had to work hard for you.

 

Hahaha. Yea I definitely gave it involves a good beating, but it's been a hard worker. Time to offer up some rewards

 

Interesting theory regarding the pasting Turbod.

I played about 5 hours of Skyrim yesterday after 4 hours of BF3, and my temps actually went up more. Shader maxed out at 89.5 degrees.

Regarding the back memory chips, do you think 1mm pads should be used here? I cant quite remember but are they like the 5870's where the components further back stop the hestsink from making a good contact? Especially if you are using a 1mm pad on them.

Just did a Furmark run for about 5 mins and temps are a little high. laptop even cut out the first time but I think it was drawing to much power. At stock CPU and the GPU's at .97 it was pulling about 290 watts. Whilst my adapter's didn't cut (cause I have two) I am wondering if the laptop shut down because of it. Maybe some else getting too hot.

I under clocked the CPU for the next run and power consumption sat around 260 watts. Temps reached 92.5 before the last shutdown, and 91.5 on the second run, but I canned it cause I knew it was just going to rise some more.

Time for a re-paste I think.

 

And the rpm of the gpu fan? over 4k?

BTW do you have eurocom card?
The (auto) fan control works?

 

i just tried download 11.11 mobility drivers from ATI and this time it really downloads 11.11 version. Lets see..

Edit: the same blue screen again just after windows logo when its ready to show desktop. maybe i should hope for 11.12 now on.

 

sorry ya if anyone ask this question before...

i install a brand "dell" 6970m at m17x r2... im getting shutdown problem. the system wont complete shutdown and stay on after the lcd power off.

anyone solve this problem before? wat is the solution ya?

 

As far as I know there is no solution for these problems.

As can be seen from the first post:

I guess I got lucky with my ES, I am not sure why I am not having these problems.

I would provide you with my vBios but I have a 6990m ES so that wouldn't work.

When I formatted I also tried the 11.11's again and it did download with the mobility auto download thing. I then had the same issue. Looks like we will be waiting on 11.12 or hopefully an 11.11 update.

 

Douse.. u still keep dell 6970m es bios? mind to send for me for testing??

 

Sorry, I have 6990m's ES so I dont have a 6970m ES vBios.

I mean, you are welcome to edit the clocks and try the 6990 one but I honestly dont know what the outcome would be.

I would not want you to brick your card :|

 

Worldedit, the problem you're having was the same issue I had with my ES card, flashing the VBIOS to another one didn't work for me.

Well, yes, the GPU fan is running at one 'notch' below the highest. It's singing pretty good. Yes, Eurocom 6990M with the X7200 VBIOS. The only thing that doesn't work perfectly is the HDMI audio out. That sucks, but the fans do control themselves.

Like I said earlier, the heatsinks are kind of POS, but with the proper attention, they are very capable of cooling the 6xxx cards with no problem. I used .5mm pads on both sets of memory chips. I can't stress the importance of getting either pads or paste on any component that is close enough to the heatsink to do so with. Every little bit counts, and I don't know how much heat those other little buggers add to the mix.

 

Are you not concerned that you are then potentially heating up other components to the temperature of the heat sink that would normally be cooler without?

If that makes sense?

EDIT:: Did a complete re-paste and haven't really noticed a temp drop. My laptop cuts power when I do a Furmark test for too long, but I think it is just drawing too much power as temps don't appear to be getting critical. The last shut down happened before the shader hit 90 degrees But, I don't have any shutdown issues in games so I am going to leave it....The more I tinker the more something may go wrong

I also did what you advised Turbod, but I pasted where you specified to use 1mm pads as those areas are as high as the component you said to use paste. I confirmed that it makes contact. I actually think using 1mm pad on the back row of memory chips may be more important...I will try that when I can be bothered.

 

Yes, it makes perfect sense, but I've already done my homework. I obtained a Clevo heatsink in a trade, which is when I truly realized that our heatsinks are kind of garbage. The Clevo sink is all copper with a much larger fin area. It has small indents in the copper for all of those different components. That is how I realized that those components should be getting cooled, when on the Alienware heatsink they're not.

 

I'm still having a similar problem, but I think it's due to heat - and bad heatsink contact.

I've had the cards undervolted to 1v and running on stock clocks for about a month now. Maybe more. I haven't had a chance to file out the slave's heatsink.

Shader temps are still skyrocketing to 106.5C under moderate load on stock voltage. Hopefully this Thanksgiving break I'll finally find the time to attack that heat sink. It's not that contact isn't being made by the heatsink - it's that not enough pressure is being applied to the die, because the magnesium (or aluminum) plate around the copper core is making contact with the square metal ring around the gpu on one side.

 

Yup, you're going to have to file down your heatsink like I did. I was really wondering how nobody had problems with the heatsink sitting on the square around the GPU die but me.

 

Hey guys,

It's very odd how some folks are having the heatsink square issues and other aren't. My temps are fantastic, and the only difference I have is that I'm using the cards with the soldered on posts. And my heatsinks seem to sit perfectly where they should. Maybe there was a batch of squares that were a tad bit higher out of manufacturing?? Have no clue, but something this whole experience has taught me is, there are a lot of odd things that make absolutely no sense

 

My temps were really good on my second card until I played a long session of BF3. Once it died, I noticed it wasn't making complete contact with the memory banks because there is a black chip that stands pretty tall and seems to keep the heatsink from touching all of the memory chips with .5mm pads. It seems my heatsink makes full contact with the dye, or at least from what I can see. The problem for me looks to be some of the surrounding chips being too tall forcing the heatsink to sit on the card unevenly.

I also tightened the heatsink down on the GPU dye pretty good. How much should the 4 screws around the dye be tightened? I made sure that it was pretty tight and the 4 screw arms were slightly bent to make contact with the posts. I wonder if maybe I tightened it down too much causing contact issues as some people have suggested. I ordered a couple new heatsinks from Ebay just to test so I don't have to keep repasting.

I'm really on the fence about having my replacement card put back in or selling it. I've lost a little faith in this project based on my experiences. It feels like I am trying to fit a square peg into a round hole. It was a great idea, but it seems that it doesn't have a very high success rate for a multitude of reasons. And even if successful, there are varying degrees of success you might experience. Might just sell these cards and put away money for a M18X R2. Whenever that may come out.

 

Also here temperatures are fine and no problem with the heatsink.

The highest reading are always the shaders that doesn't pass 77-79 degrees on both cards. Max for the other sensors is around 73 -74 degrees. And this are the spikes of temps, as soon as the fans kicks in at full speed (triggered by the shader sensor passing 75 degrees) they drop under 70 but the shaders that remains at around 72-74 degrees.

 

Cheers Turbod, that is what I want to hear! I don't mind cooling these bits now

Ghost, I know exactly what you mean. I think I am making contact with the GPU die as the thermal paste is always spread across the whole surface when I remove the heat sink. The problem that I see is also similar to the 5870's. The back row of memory chips is being stopped from making good contact and I can see this as the thermal pads still have that checkered pattern across 50% nearest to the tall back chips....so only half contact.

Ghost I dont think over tightening it would have caused your issues, the metal square would stop too much pressure from being added to the die and the memory chips are quite robust.

I am going to put 1mm pads on these and see how I go....or I might whack a big dollop of X23 like I did with the 5870s

Speaking of oddities, I sometimes get graphical artifacts on my second monitor when showing temps etc in a game. I am sure this is not a heat issue as it sometimes does it immediately when starting the game so I think it is drivers. Other times it doesn't do it all and seems to happen more with Skyrim.

Turbod, I am skeptical about the filing of the heat sink as I do not want to add undue pressure to the die...I mean how fail can the video card manufacturer's be if they don't make the metal square around the die exactly level with the die....would absolutely defeat the purpose.

Greywolf, I am also shocked at how many things vary from person to person. I personally think it is because there are so many brands of cards flying around, each having it's own little quirks.

 

i tried flash to clevo vbios... the system can be shutdown but display fully with vertical line and alot greenish dot.

i think no chance for "dell" card work with m17x r2 already.