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

Stick the tube near a light bulb and it should be easier to spread, or stick it in a glass of hot water. I do that with ICD if I need to use it. But it does sound like doubling up may help out a bit. Here's hoping!

 

Yeah I place the tubes over a lamp and they get pretty hot, but even then it is still a pain as you can't just use the dollop method...you have to painfully spread it out.

I am thinking of spreading it with some glad wrap or something.

How do you apply it?

 

I use the heat method with a flimsy piece of plastic thats kinda stiff, like the plastic from the back of the EK thermal pads I bought. Though the Glad wrap may work even better. I'll have to give that a shot next time I repaste with ICD

 

1mm pads on both sides , ic diamond (heated up in a cup with boiled water) on DIE.

 

Ok bugger it, I am doing it now.

I will post results in a bit

 

/sigh

Well that was not very successful.

I used 1mm pads (2 x 0.5mm) on the memory chips, X23 on the Die and 0.5mm on the components around the memory and temps are worse then before.

I even got a thermal shutdown playing Skyrim and there is no way I can run Furmark.

It is that damn shader and the memory which are the problems.

I did another repaste of the die using the line method with the X23 and it appears to be a little better but still way to hot....in fact my first attempt using TX-3 seems to have been more successful.

At this point I am at a loss, I have repasted 5 times and each time results are similar so I do not think it is a bad paste job each time.

Maybe the heatsink is not making proper contact as you have suggested Turbod. Time to buy a file me thinks.

Anyway, here are my current results. This is after about 30 mins of Skyrim. Depressing I know.

EDIT:: Similar results with BF3, the mem and shader a full 10 degrees more then the the other two sensors...

 

You shouldn't need more than .5mm on the VRAM chips. When you use 1mm, all you are doing is keeping the sink from contacting the die just a little bit more. I'll bet your heatsink is sitting on that square around the GPU die. I run a 800/1000 full-time overclock, and I can't even hit 80°C while gaming on a single GPU, which gets much hotter than a CrossFire setup. Check for the heatsink sitting on the metal square around the GPU die, and you should be using .5mm pads, or you are just worsening your problems.

 

Hey Douse,

Maybe I missed it, but have you tried undervolting your cards? I'm using OCZ freeze (spread method using a plastic sandwhich bag), and .5mm (although I've tried 1mm) fujipoly thermal pads on the memory only (not on the bottom of the card), and my Eurocoms rarely go above 65C.

I haven't run Skyrim on my system (got it for Xbox but that's another story), so can't do a direct comparison. I've only run Vantage, 3dmark11, and Shogun Total War 2 benchmarks.

I'll see what the temps look like running Metro 2033.

I realize you are running 6990's, which probably run hotter, but I can't explain the difference being that significant.

Either I'm the worlds best paster (ha, although I do have a TON more experience now), I have some strange cooling going on (I am using a Cryo Pad, but ambient temps are still in the mid 70'sF), or something either in my card is not being reported properly or some other anomoly is going on.

I don't know if I feel comfortable that my temps are correct, or if something else is going on.

So far no screen corruption or shutdowns, and my kill-a-watt rarely reads above 240 except in benchmarking.

Hope you can get it sorted out though, I don't think its your paste job, it's gotta be something else.

 

Thanks for the input guys.

Greywolf, I am running the cards at 0.97 volts but I dont know if there is a way I can check this?
Running 3dmark11 does not spike my temps that much, I am about to test Skyrim and see if there is a difference.

Turbod, I came to that conclusion as well so despite being ready to call it a night I gave it another tear down and this time used X23 on the memory chips (like my 5870's), making sure to put an extra amount on the back ones.

Being that I used so much X23 I had to use TX-3 on the die.

I figured, this will allow the heat sink to sit as low as possible and provide me with the best possible contact with the die.

I only did this on the primary card.

Results are much better. Yes, Furmark still gave me high temps and i did get a thermal shut down but I think that is because my second card was hotter then the primary card and it took A LOT longer to get there.

Before I had to kill Furmark within a few minutes, this time I think I left for about 10 minutes before temps started to get a bit shaky.

I will have to order some more X23 for the other card so I can replicate it but I dont think I will have any more problems as games are never going to pwn the system as hard as Furmark.

EDIT:: Well did some game (BF3) testing and very happy with the results. My primary card is running cooler then my secondary card, which is saying something. Before i played some BF3 and the primary card was reaching 90 degrees easily. Now it is happy sitting around 80. See screen shot. You can also see that there is not as big a gap between all the sensors.

EDIT x 2 :: Just played the same segment of Skyrim, and primary card sat around 81 degrees on the shader. Secondary was about 5 degree's hotter. Hopefully these drop a little as well as X23 has a tendency to need to settle in a bit.

 

Awesome Douse,

Looking much better.

I'm always the same way, say I'll leave it for the night, and then tear it down...again.

 

Thanks Greywolf!

Yeah, definitely gonna hit the hay now though!

Your temps are seriously awesome! You must be stoked! I wouldn't touch those cards at all

So does anyone know if there is an app that will give me the cards voltages? I know I cant just use atiflash to export out a copy of the firmware but it would be nice to get confirmation. I didn't really notice a change in temps when I dropped the voltage.

 

Congrats on the temps. That's where mine are. They should be perfect when you get both cards done. Out of curiosity you should disable crossfire and run Furmark. See where that primary settles in at.

 

use this thing bro
and in options tip everything on, it'll show you real time load/fan speed/voltage
you can run 2 of those at the same time, set first widget to minitor master card and second widget to monitor slave card

Also, hommies
for some of you 0.5mm pads are fine, for me not, i took a closer look on my heatsink and it is placed perfectly on the metal square that protects DIE. The gap between mem chips and heatsink is twice as big as 0.5mm pads. I would even say that for perfect squize-in i would need 1.2mm pads, therefore i assume, think, and i am almost sure, our heatsinks are not the same.

Or i don't know, maybe USA versions are a bit dif than UK ones.

 

GPU Observer is cool, but I wish it would show both GPUs Am I missing something?

 

Thanks flingin! However the option to read VPU load is greyed out and the value I get for fans etc are just stuck on 30% which is the same as HWINFO.

DR650, I disabled crossfire and ran Furmark on full screen, extreme Burn-in for about 10 minutes.

It is still a little hot, but it takes a while to get there. It creeps up 0.5 of a degree every 15 seconds or so. Almost like the heat sink is at its limit with the card. I think now is the time to make it more efficient if that is possible...

It seems that the card never settles though.....sllooooowwwweelly keeps increasing just a little at a time..

Anyway here is a screen if people are interested

EDIT:: Did another Furmark run with the Burn-In disabled. Lasted for about 20 mins with the temps stable at 94/92 on the mem and shader respectively. But then had a thermal shutdown....I dont think it is the card causing it as those temps were stable...something else must be causing it. Going to take off one of the back pads as it makes contact with the motherboard and it may be leaking heat in to the system.

Oh and i found that Furmark actually reads voltage lol! It is sitting at 0.97

 

You have to run 2 gpu observers and in options change video adapters, i dont know if there is any gpu obs designed to monitor 2 cards simultaneously :/

 

Hmmmm I am thermal shutting down like there is no tomorrow lol.

I'm gonna have to check the paste and pads. Idle temps are around 50. Everything runs fine until I load up any game. Runs for a few minutes then BOOM it just cuts off as if somebody just came along and pulled the AC Adapter out and there was no battery in use either.

Cheers.

 

Yep, thats the same as what I was experiencing.

I am fairly confident that I wont get it in games any more as a 20 minute run in Furmark is pretty good in my opinion. Wouldn't last more then a couple of minutes before.

 

Awesome man! I'll need to do the same as you and hope for the best.
A bit too hot and lazy atm lol.

Cheers.

 

Yeah it is a pain but worth it. I was getting a bit frustrated when every thing that I did was making it worse

Honestly, I think the 6990m's are pushing the limits of the heatsink.

I am open to any suggestions to improve them. Maybe fan upgrades? Up the RPM?

If there are still problems, I will be following Turbod's advice and filing down the surrounding area around the die.

 

when your card is just fine in idle, but temps skyrocketing under load, then you have a contact issue between die and heatsink for sure.

Leap it

And it should be fine

 

Yeah, I agree. They are not skyrocketing as such any more, it is just a really slow increase, almost like the heatsink is being saturated and not being able to get rid of the heat quickly enough.

 

this is wierd, i feel like my 6970m is running cooler than 5870m, both under load and idle.

Maybe you should change the termal paste to icd 7 as this one is thick and i feel it would make a dif in your case.

 

I have tried both X23 (which is really thick) and TX-3 (which is quite thin) and results were actually better with TX-3.

I might give it a shot. I expected the 6990m to be hot and apparantly the shader temps are reknown to be a bit of an issue.

 

I just tried to set 700 on gpu core while i am still on 0.96v

And it worked, 3dmark 11 P3300

temps still max 76 (shader)

Anyone tried to combine OC/UV ??

 

Woah, 3300?

Seems a little low!?

Oh sorry thought you had crossfire! My bad

Have you run Furmark with your card? Mine sit around your temperatures after a 3dmark11 run. It is just Furmark that kills me.

 

There is nothing wrong with the 5870 heatsink. It can more than cool the 6990. I'm running a single card, and that makes it cook a bit more even. You have a contact problem somewhere.

I am willing to bet that you're sitting on the metal square around the GPU die, OR you're sitting on the one voltage regulator right where the two sets of VRAM come together. You can re-paste and re-pad all day long, it isn't going to do any good. And no, you don't need to lap your heatsink, that isn't what I am talking about. You need to get a small metal file and file down around all four sides of the copper plate on the heatsink. This will make it so that the heatsink is actually sitting firmly on the GPU die, instead of sharing some of the compression with the metal square there on the card. Might as well just get it over with. Trust me.

I run an 800/1000 overclock all of the time, and my card never goes above 77°C, no matter how long I play.

 

I know what you are saying Turbod, and I may do this. However if I can avoid it I will.

At this rate I should be able to get both cards within respectable temperatures. When I get bored and I need to overclock I will go down this path.

If however I cant get the temps under control I will certainly take your advice.

The only thing I am concerned about it placing too much pressure on the die.

 

Douse, I been reading your temp lowering saga with interest, sounds like you exhausting all options except TurboD's advice, it is probably time to get that file and give that a shot, it should work for you as well if it worked for TurboD.
If you decide to give the file a go, please share some pics of the process with us if possible, those would be a great addition to this thread
Good luck on that!

 

I suppose it is looking that way....

I suppose I mangled the back plate anyway, whats a little more modification on the heat sink

If it needs to be returned to stock, replacement heat sinks would have to purchased anyway..

Haha me and my rubber arm!

I thought I would do a comparison for you so I ran a 3dmark11 run with a single 6990m and overclocked 920xm. I scored ~3500 so you are not far off a 6990!

 

That's nice man
At the moment skyrim runs >30fps constant on full settings, other games as well.
I will test how far i can go with 0.96v, i belive it is almost max (700) though.

These cards seem to be way overvolted to me, honestly, if i can run it at 0.96v and still do some oc - what more could i want

 

Yeah that is pretty awesome! Even if you have to up to 1.0 volt its no biggy.

You know you could easily run the stock voltage since you only have one card yeah?

Should be able to handle even an overvolt with the CPU fully clocked up!

BTW I am fairly happy with my temps actually. This is after 1.5 hours of Skyrim

 

Flingin, this on a single6790 right? if so, thats awesome, I am still on my xfire 4870s playing Skyrim on HIGH settings, not Ultra, and I love the looks of the game, must be great on Ultra

 

Douse, what have you done to the heatsinks that causes you to think that they need to be replaced? I'm sure they're fine. Here are the spots that I filed down. Your temps are too high for a CrossFire system; they should be more around the 75° range. Get a small flat file and work around the square copper plate. I guarantee your temps will drop.

Where to file.

This is one of my hacked up heatsinks, but the principle is still the same.

 

Douse, nice Drizzt picture ^^

 

Thanks @tomX!

Turbod, I would have to buy replacement heat sinks because I chopped the back plate for the installation of the 6990m.

Therefore if I were to ever log a warranty call I would have to revert it to stock including the back plate which would come with a replacement heat sink...

I have noticed that the components behind the memory chips actually sit flush on the heatsink. This is why I am using paste on them.

What if these are what is causing my heat sink to make improper contact. I believe it is these that are causing the 0.5mm pad to be not enough on the parallel row of memory chips...but 1mm to be too much. That is why the X23 is perfect because it allows it to be the perfect fit.

 

Great read and info provided in the last few pages. Thanks everyone.
I finally decided to give it another go. Re arranged my pads to make sure everything was touching and re-pasted. Not freezing anymore or thermal shutting down. So that is a good start.

I managed to run a quick 3d mark 11 run.

AMD Radeon HD 6970M / 6990M video card benchmark result - Intel(R) Core(TM) CPU X 920 @ 2.00GHz,Alienware score: P3397 3DMarks

Bench run with all settings and hardware on stock settings. I'll do some more testing soon with CPU OC and GPU OC + CCC Tessellation set to OFF.

Tessellation set too OFF = AMD Radeon HD 6970M / 6990M video card benchmark result - Intel(R) Core(TM) CPU X 920 @ 2.00GHz,Alienware score: P3797 3DMarks

BRING ON the M17x R3 lol! = AMD Radeon HD 6970M / 6990M video card benchmark result - Intel(R) Core(TM) CPU X 920 @ 2.00GHz,Alienware score: P4124 3DMarks

Cheers.

 

Glad to hear it. I am getting similar temperatures as well.

Nice scores Chief!! So do you get a boost by turning tesselation off?

Going to swing past a hardware store on the way home from work and pick up a file (Most of my tools are still back in Aus )

Since I am resonably happy with the primary cards temperatures, but was going to repaste the secondary anyway I am going to attempt the filing method only on this card

If it is successful then I will then do the same on the primary. It will be a good comparison imo!

 

Hey thanks Douse man!

Yeh for AMD GPUs you can get quite a boost if you set the Tesselation to off in the AMD CCC. However it is illegal to do so if you want to register your score on 3d mark 11s webpage lol.

It's time for me to push on and try to catch some of the M17x R3 scores lol!

Hey DR, Are you able to setup a benchmark part in your top 5 for M17x R2 single GPU top 5 scores?
If so that would be awesome lol! TIA!

Good luck with your trip to the hardware store.

Cheers.

 

Nice scores man, I might consider creating another catagory when I get more time. Might help inspire some more traffic in the bench thread. But it would be another catagory I'd have to dominate it

Vantage: Odd how I set the GPU clocks to 840/1080 and the max clocks on HWInfo showed up at 865/1075. Wonder how fast the card was actually running. Never noticed that till now.

Spoiler

3DMark11: Done with Tesselation off, before Futuremark update on tesselation feature. Still could have pushed higher, but was benching everything just to get through them, all. Didn't really push to much on the single card benches.

Spoiler

 

Well, picked up a file set from the hardware store but the missus wants to head out for food and drinks....

So will have to wait till tomorrow...

 

HAHA! Nice one DR!!!
Your scores look like they rival or are really close to the lowest top 5 M17x R3 scores man!

You've got me beat already LOL as always ahahahah!
It's strange and so annoying though, atm I can push the clocks much higher but the benches keep dropping out and crashing to the desktop and then showing that error message saying like the driver has crashed or some other error has occured. You guys know, the error like if someone hits the ESC key during a bench for some odd reason. I would much rather see it freeze or thermal shutdown instead of exiting and crashing the bench!

Any ideas peeps?

I'll try and give your clocks a go too. Also strangely enough 3d mark 06 has disappeared from all my systems LOL! Poof it just uninstalled itself. I'm gonna have to reinstall.

Cheers.

 

I think it's just lousy AMD drivers, reborn. Some systems seem to be affected more adversely than others using the exact same driver for some reason.

 

my question is they didnt receive negative feedbacks? what did they reply to those feedbacks? and as i can see its not a few users issue. And finally why they didnt prevent 11.11 drivers from being released as causing so many issues or even useless at all?

 

True that Fox,

Just ran a couple more quick benches. Still no where near DR's scores however slowly creeping closer. I should be 2nd at least on the top 5 list for now LOL! I need to increase my CPU score in Vantage, thats where DR destroys me and leaves me for dead. I've got no idea how to use TS properly though. I just click here click there and hope for the best.

I can't wait to get a couple of HD 7000M GPUs to test
I'm keeping my eyes peeled this time around and will ponce at the 1st chance I get to purchase them. Hopefully this years Christmas bonus is enough to cover them.

Cheers.

 

Hey Everyone,

So I went ahead with the the heat sink modification and I have not noticed any difference with my temperatures.

I reverted back to 0.5mm pads on the memory and paste on the die and components.

Re-pasted twice since the mod with similar results.

So far the best seems to be the X23 on the memory chips, as I have done on the primary card as the secondary temps are often 7-10 degrees more (shader IO).

I have found a very strange and frustrating thing which may be causing my high temps. I noticed in Furmark that for some reason my voltage is 1.1v.

I have re-flashed numerous times and created new flash files etc but the voltage is always read at 1.1v. I am sure it is actually using that much voltage because I can see it in my watt meter.

So I am at a bit of a loss now. I dont see how it could be running at 1.1volts when the firmware is telling it otherwise. Furthermore why did it even revert!? I KNOW it was reading a 0.97 the other day....

Either way I did take some pics of the process, and here they are.

EDIT:: SO, this time I altered the BIOS to run at 720/900. Did a flash and even -f it. No errors, but then running Furmark, clocks are at 715/900...dont know what is going on

EDIT x 2:: ARRRGG I HATE OVERDRIVE! I had not even gone in to it and it causes me issues! @#&@#*&#@^*&@#*^@#$. So i enabled overdrive and disabled overdrive and LO AND BEHOLD the card is dong what it should be doing.

 

Douse, I apologize if it seems like I've sent you on a wild goose chase. When your heatsink is fully installed and tightened down, and you look at the sink/card from the side, it the card slightly bent anywhere? I noticed it was really bad toward the back of the card, opposite side of the die. This was without the retention mod or anything else...normal compression. So then I actually went and filed down more of the sink, which I'm sure is the last thing you want to be doing. You don't want the heatsink to actually be sitting on those other components. Most of the compression should be on the GPU die, the VRAM should be getting the rest. I had to file...and file...and file, until just a a tiny bit of light shown through between the sink and those voltage regulators. Then I used paste or pads where applicable. My card and sink are perfectly parallel, if that makes sense.

 

No no, you haven't sent me on any goose chase!!

I know that the heat sink is actually sitting on this regulators at the back, so that will be the next step....

My computer is unstable for some reason and I think it is overdrive so I have rolled back the image and see how it goes.

EDIT:: Ok, so I have gone file crazy and while it has made a small difference, my second card is still running a lot hotter then my primary card. When holding the heatsink on the card it certainly rests on the regulator at the back. I think I am going to have to file it down a lot for the memory chips to be flush to the heatsink. Also the area that is filed is quite coarse....

Had enough for today though. Might just order some X23..or a bigger file

 

I am really struggling with overdrive. I am convinced it is causing my system to be unstable but when I roll back my system the cards do not run at the altered voltage until I ENABLE overdrive......

Is there a way that I can uninstall overdrive? Or install drivers that do not have overdrive incorporated?

 

Hey Douse,

I've never had Overdrive as an option in my CCC until I recently uninstall the ATi CCC whole drive package and then I ran driver sweeper in safe mode. I then installed the Preview 12.1 driver set from these forums. Then there was overdrive in the CCC performance section. I haven't used it as I normally use TRIXX to OC.

So you may have to uninstall all your drivers or at least the CCC part and then reinstall with a different or older version which does not include overdrive. It may also be a registry thing that you need to uninstall or remove.

Cheers.