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

The 5870 heatsink needs to be slightly 'modified' to work on the 6xxx cards sometimes. There's a good chance that your heatsink is sitting more on the square deal surrounding the GPU die than the die itself. I had to file my heatsink down to get it work.

 

Oh yeah you also need to turn off ULPS otherwise HWINFO32 does not detect the secondary card when idle. Normally it wouldn't be such a problem, but because we rely on the application to manage the fan speed we need to detect upon opening, every time.

Pretty much open regedit and go to: [HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Control\Class\

Search for ULPS and for every one that you find that says 'enableULPS', make sure you change the registry value to 0. You should find around 4.

 

I thought I would put up a post regarding my throttling experiences of late to see if anyone can make sense of it.

As some of you may know, I re-pasted my CPU a few days ago. However, I also used 1mm pads instead of the 0.5mm pads as I suspected the heatsink was not making proper contact to whatever it is next to the CPU (voltage regs??)

Since I have done this, not only are my CPU temperatures lower and more stable but for some reason I get much, MUCH less power throttling. I have been able to run BF3 with an overclock on both the CPU and GPU with no throttling.

Power draw can be up to 280/290 without any throttling (haven't pushed harder), and I have been testing for quite a while. Ambient temp is the same as usual because of the central heating.

Previously if I went over 270 I would risk a throttle.

I did not touch the video cards at all during the CPU re-paste. The only other thing that I did was clean all the fan grills etc, however I do this all the time.

So, interesting that this seems to have had a direct affect on the amount that my laptop can handle in terms of power draw.

Perhaps those components next to the processor are involved with power throttling? I honestly don't know.

Anyway, I will obviously do some more gaming and post back if anything changes.

 

Where are you refering to in terms of a 1mm pad? There's only that one little raised nook on the heatsink that a pad has a place for. Is that where you are refering to?

 

When you take off the CPU heatsink, there is a stock thermal pad which sits across the heat pipes and touches some components next to the CPU.

I originally put 0.5mm but I found that it wasn't really touching. So I replaced that with 1.0mm when I re-pasted with IC 24 carat.

A little update, I ran BF3 with a full overclock on my CPU (23x all cores) and it happily sat above 290 watts without throttling for hours. I dont know what I have done, but I am not complaining!!!! This means I can run the CPU at stock and overclock the crap out of the GPU :GEEK:

 

I don't have BF3 so I don't know how mine does, but I think I probably have a 1mm pad on my CPU heatsink.

Can you do me a favor? Can you take off ALL of your overclocks and extra power settings, like you're running @ stock, and tell me what your CPU vantage score is? I want to make sure 15,000 is normal, and if its not - why.

On a side note - what do you think about disabling virtual memory file paging?

 

The components next to the cpu are voltage regulator modules. They control the voltage (and the quality of the voltage) for the cpu. If they're overheating then it makes perfect sense that they could cause throttling.

 

That's odd. So you put the thermal pad on the components themselves when you laid down the heatsink? Or did you put it on that raised part. I'm pretty sure I put a thermal pad down on it when I redid my CPU. I asked greywolf about it when I was doing it and I'm not sure if he knew what it was, or whether or not it was nescessary to put a thermal pad on it. I can't remember though. I'm sure I'd be having massive performance issues if there was no thermal pad there at all, right? Also, aarpcard - can you run your cpu @ stock clocks / power settings with no OC and tell me what you get for you vantage score (CPU)? I really would like to know whether or not 15000 is adequate for stock clocked CPU, and if not - why mine isn't where it should be.

 

Here you go Intense:

AMD Radeon HD 6970M / 6990M video card benchmark result - Intel(R) Core(TM) CPU X 920 @ 2.00GHz,Alienware score: P19891 3DMarks

I did a Vantage run with stock CPU and GPU clocks. The cards are set to 1.03 volts.

 

Perfect! I actually have a higher score than you with stock clocks

I'm glad to know everything I have is working fine. All my ingame performance is perfect now. I got everything in SC2 situated. What I thought was abnormal was, in fact, normal. Reformatting + new xfire cable + new psu worked beautifully. Plugging in the old PSU actually throttles me performance, so it was a good investment.

http://3dmark.com/3dmv/3809949 not sure why it says both of our ram is 667mhz

 

There you go

Because the RAM is running in dual channel. 667MHz x 2 = 1334MHz

I suppose that would make sense if I was pushing my CPU, but I had it running stock. Perhaps those regulators aren't just restricted to the CPU?

 

Where on this picture is the power thingy that the pads need to be over? I got this image from a teardown video. Just curious because in the video he has the pad on the heatsink but there is still more metal that (i can't remember if) more pad can stick to. I'm just trying to determine how much of that metal piece needs to be covered with the pad. Is it correct in the picture? Here is the link to the video. The CPU heatsink comes off @ 7:35. Alienware M17X R2 CPU upgrade from i5 Dual Core to i7 Quad Core engineering sample - YouTube

 

It is exactly where the pad is in the second photo. If you were to take off the CPU heat sink it would become clear. Not that I recommend that if everything is finally running well

There are those 3 voltage regulators in a row, so the strip is about a 1cm wide and perhaps 3 cm long as seen in the first photo.

All I did was use a 1mm thickness instead of 0.5mm thickness.

There is nothing special about it, just make sure it covers the 3 regulators....

 

When I replaced my CPU that was the one thing I hadn't done and now wish I had (The thermal pad for the CPU). Since everything is running fine, I don't plan to take it apart just to replace it...just wish I had.


Thanks Intense for the photos!

 

I'm confused. Is there a metal piece indicating to put a thermal pad there? I can't remember if I did now. Looking at the photo here, it looks like its beneath the metal piece where a thermal pad should go (seemingly). Now I'm really even more confused. I'm not sure if I would have put a thermal pad just on the copper heatsink.

 

I dont really understand what is confusing you

It is exactly like the picture. Yes there is space for the pad to go and you wont really be able to put it anywhere else otherwise it wont make contact with the components....

 

I remember now. Sorry. For a second I thought it was just the copper pipes that were under that pad in the picture, but now I remember there's a flat piece of metal for the pad there. Right?

 

Actually, from memory, it is directly on the copper heat pipes, with a bit of extra metal (the grey metal) on the sides to give it the width. You can kind of see it in your previous picture.

It doesn't really need a polished plate as it does not need to remove as much heat as efficiently as it would on a die.

 

Eh, hopefully I put it on there. I have good CPU temps so I hate to open it up and recheck it. I'm sure I'd notice performance issues if I were missing it, right?

EDIT: I have a question, something I'm concerned about. My cards are running @ stock voltages because, for whatever reason, overdrive enabled makes my ingame performance smoother and it keeps the cards @ stock voltages. My temps never ever get past 80, and in SC2 they hover around 72-73 tops. I know one of the major reasons to undervolt is to get lower temps, and lower temps promote longevity, but with these temps do I need to worry about my cards frying any time soon? I really need these cards to last a good year or so. Think, given my temps, they will? With no 3dclocks running they stay around 40-45C.

 

Yep, if it ain't broke dont fix it

Umm, that isn't entirely correct.
The main reason that people under volt is to stop the laptop from throttling. Less voltage = less power draw = less total power consumption. The drop in temperature is just a good side effect

If the cards were overheating with stock voltage then I would say there would be something wrong with the installation of the heat sink, which would need to be addressed.

Your temps are fine, and I would not ever worry about the lifespan if your card unless you were frequently getting 100+ degrees and thermal shutdowns. Even then, they are operating within their limits as this is what a thermal shutdown is designed to do.

Also, I think you may have misunderstood the Overdrive issue that I had. Overdrive does not have an effect on your set voltages. It ONLY affects clocks. Well, it should only affect clocks.

In my case, unless I had ENABLED overdrive (by ticking the T&Cs) my voltages remained stock. So for what ever reason, my firmware was not being applied unless I had activated overdrive.

I am sure that this was a once off thing with my ES cards, so you should be able to use overdrive AND under volt your cards with atiflash.

I am currently running an under volt that was applied with a firmware update, and I am using Overdrive to overclock my cards. To confirm, I just quickly run Furmark to check my 3D voltages and also my 3D clocks.

 

Hi guys,

Had this project running myself for a while, but have had to put it on the back burner for a while due to work commitments. Quick version: Had 4870's in crossfire, and upgraded to 6970's. I suffered similar problems to greywolf in that i could not get crossfire working. This was around 4 months ago.
I sourced an m18x cable the other day and finally booted the laptop up yesterday with the new cable and Bingo! crossfire enabled!!!!

However now hit my next problem. Im experiencing massive performance drops in games. From like 60fps straight down to 10-15fps in most games Usually as soon as the are running. I looked at the logs on gpuz and it seems the cards are down clocking(to 150mhz for example) for a few seconds then firing back up. I presume this is what you guys refer to as throttling??(sorry im a noob at some of this stuff).
Its not due to the temps as far as i can see, they seem fine, (ive never seen them above 80). I was wondering if anybody else has had this problem or would anybody be kind enough to give me a few pointers? and things to try??

O before i forget I've got a 920x cpu and im running the 12.1 preview drivers. Im turning it in for the night but im going to crack on in the morning and see if i can figure out whats going on. So thanks in advance for any suggestions or replies

 

According to GPU shark, with overdrive enabled my voltages go to stock. Even if I have an undervolt, for whatever reason, the voltages go back to stock. Not sure why.

 

I'll try it tonight with my cards.. if it is like this, i'll call it a feature (at least for me), being able to keep cards undervolted at stock, and bump the voltage up as soon as OC is applied is not that bad at all

 

Wow, that's crazy that we have completely different results with overdrive. CptX, let us know how you go!

I NEED overdrive to be enabled for my voltage to drop! Strange!

BTW, I did the 5% FSB overclock and I scored P23345 in Vantage, up from P22619

AMD Radeon HD 6970M / 6990M video card benchmark result - Intel(R) Core(TM) CPU X 920 @ 2.00GHz,Alienware score: P23345 3DMarks

EDIT:: Well, this is strange. I double checked my results and with overdrive enabled my primary card is at the correct voltage. But it seems that my secondary card goes to stock....arg. Furthermore, I am getting conflicting information between Furmark and HWINFO32. My cards are clocked at 740/940 and this is shown correctly in Furmark, but HWINFO32 shows 740/925...

Ah bugger it, it runs fine

 

Are you running any other OC's?

 

To get that score? Yeah, my cards are running at 740/940. Well, not according to HWINFO32 which shows 740/925....

 

What about CPU?

EDIT: In your former vantage test ( http://3dmark.com/3dmv/3819164) your CPU clock was @ 3192. In your newer test, your clock was @ 3762. Was that just the 5%? I'm not sure why, but for whatever reason, the 5% OC made my performance worse in games. No clue why it did that.

 

Using throttlestop, the same as before in my benches. 24x across all cores, 25x for dual core and 26x for single.

Nah, that test that you referenced was my completely stock system. So this is with throttlestop and the 5% overclock. A better comparison would be with AMD Radeon HD 6970M / 6990M video card benchmark result - Intel(R) Core(TM) CPU X 920 @ 2.00GHz,Alienware score: P22619 3DMarks

Hi Clarky,

Congrats on the install and getting crossfire enabled!!

Your issue does sound like throttling, although I haven't noticed whether the GPU clocks change during a throttle so I can't confirm if that is normal. From memory, I didn't think that they did

I am a little unsure though being that you have 6970's....you shouldn't experience that much throttling unless your CPU is overclocked.

However, if you are, the best way to find out is to buy a watt meter from your hardware store and you will see the power consumption peak around 260/270 and then drop in correlation with performance degradation.

In order to get around this, members have under volted their video cards in order to reduce the total amount of power that the laptop uses. Some people have been successful running at 0.96 volts, however this is dependent on your cards. Others haven't really been able to go under 1 volt.

If you dont want to spend the money on a watt meter (which is fair enough ), just go ahead with the under volt and see if it makes a difference as you can always flash it back if need be.

 

Gotcha. Does that really make a difference in games that you play? I keep everything @ stock. No Ocing whatsoever.

Why do you think the 5% FSB would decrease my performance?

 

Hi Clarky, sounds like the issue I had with the GTX 260 on the M15X( http://forum.notebookreview.com/ali...tock-clocks-overclocked-171.html#post6535976).
Throttling provoked the graphs in GPUZ to shape a regular pattern of ups & downs from max clocks to minimum and also, as mentionned by Douse, the reading on the wattmeter to jump from 150 down to 80 everytime the graph hit the lowest clocks.

 

I think it does make a performance difference in games, although nothing mind blowing.

I am not sure really. My first thought would be that your CPU was overheating due to the extra clock but we know that isn't the case.

TBH I am not sure

How do you know that you have a decrease in performance? Have you done a comparison benchmark?

Anyway, my GF is nagging me to go to bed, so I'm off for the night!

 

Hi Clarky,

I agree with what Douse has suggested. Try undervolting your cards and see if the problem goes away. I personally haven't had any throttling, but then again I've been running undervolted at .96 almost since the beginning.

Also check your temps...make sure they aren't heating excessively which could be causing throttling as well.

Glad to hear the m18x crossfire cable worked for you...I've been recommending that everyone who plans to go this route look at getting a crossfire cable for an m18x or from Eurocom...as there have been a few instances where the r2 crossfire cables just don't work (although I've yet to understand why its so inconsistent).

Good luck to you.

 

Haven't done any benchmarking tests but I do notice that the game just doesn't feel as "smooth". Things work just fine now. I'm sure I can get a higher benchmark score or even better skyrim performance but I don't play demanding enough games to really warrant risking a problem. SC2 works fine, but of course I'd love it to work even better in mass combat situations.

Do you OC at all grey? I can't remember.

 

I just had a question kinda curious.......does it make a difference plugging the AC adapter into a surge protector/power strip vs. the actual outlet? Would it draw less than its capable on a surge protector? Im gonna take a guess and say no and theres no difference besides the surge protector being safer?

 

Hi Pewpew,

I can't say that I've seen any difference. I have mine plugged into a surge protector (as well as running a kill-a-watt meter) and I can't say that I've seen any difference in power draw between the surge protector or the outlet itself. Maybe aarpcard will know of one, but in my experience/tests I haven't seen any difference.

@Intense...just running the 5% OC from the Bios.

 

okay thanks

 

So, I reinstalled the driver (12.1) and was able to get my undervolt to work with overdrive. After I reformatted I remember the undervolt working with overdrive until I disabled it and then reenabled. Now that I know I want to keep it turned on, I don't need to futz with it. After reinstalling the driver I accepted the TOS for overdrive and it was enabled with my undervolt taking place. I also enabled the 5% BIOS OC and I am noticing a nice little performance boost. I guess before, my power would throttle because it can't handle the 5% OC @ stock voltages. I'm running @ 1.02V with the 5% OC and everything seems to be great.

 

There will be no difference in power draw. They only problems you might face is increased electrical noise if you have a bad surge protector.

 

So I have the 5% OC enabled with the 1.02 undervolt working.

AMD Radeon HD 6970M / 6990M video card benchmark result - Intel Core i7-950 Processor,Alienware score: P20690 3DMarks
AMD Radeon HD 6970M / 6990M video card benchmark result - Intel Core i7-950 Processor,Alienware score: P21283 3DMarks
AMD Radeon HD 6970M / 6990M video card benchmark result - Intel Core i7-950 Processor,Alienware score: P20764 3DMarks
AMD Radeon HD 6970M / 6990M video card benchmark result - Intel Core i7-950 Processor,Alienware score: P20945 3DMarks

For whatever reason my clock speeds are different in each test. Any idea why?

EDIT: Also, according to hwinfo64, it says my TDP and TCD are both 62. One time I enabled throttlestop in order to test something and I know those are the numbers it has in there by default. Would running it one time save those variables permanently? Should they be 62? If now, what should they be so I can plug it into throttlestop and resave?

 

Hey guys, I have a few questions that my searches have not gotten any answers to.

Does the Clevo 6990M work in an M18x by any chance?

And, does the Dell 560M work in the M17x R2?

There's a pretty stellar deal on the Dell Outlet on an M18x, but it only has the 560M, and I'd like to do a swap. Any help is greatly appreciated.

 

I remember Aikimox installing two Clevo AMD 6970Ms in the M18x but the system couldn't detect the 2nd GPU. There's a thread about it too:

http://forum.notebookreview.com/alienware-m18x/605198-clevo-m6990m-cf-m18x.html

As for a Dell 560M working in the R2, I'm going to say it will since I tried both a GTX 460M and I currently have a 580M, but there would likely be fan issues (i.e. fan constantly stuck at one speed). Also, they have tried a 560M in the M15x and it works.

There seems to be some good news about 7000M compatibility in that CES 2012 featured a Clevo X7200 running Crossfire high end 28nm mobile cards.

And the 580Ms should be down in price by now, at least I would think given the price drops from systems equipped with the setup.

 

Hi thanks for the reply, looks like i'll try a small under volt see if that helps then. I'll just do a bit more investigation first then go down this route. A quick look at gpu-z shows my clocks spiking regularly like somebody else mentioned,although my temps do seem fine. Never had any experience undervolting, Ati tray tools can do this as a temporary measure to see if it works correct?

The cable seems good, if a little short but nothing a little electrical tape doesnt sort out thanks again for the replies guys i'll report back later with some progress .if i can figure out this undervolting that is!

 

That is strange regarding your TDP/TDC Values.

It wouldn't surprise me if this was the case, I have had throttlestop set stuff even when it was not open.

I would recommend installing Throttlestop again, and hit the 'turn off' and 'turn on' button a few times. Then once it is turned off (the button should say 'turn on') hit the save button and close Throttlestop.

This may explain your slight performance increase when we compared stock runs.

I totally agree, I wont be swapping out these puppies for a long time. They perform great and probably will do for a long time to come.

 

Thank you for that information, Ichime. The system won't let me rep you right now, but I will once I can. Thanks again.

 

Throttlestop was never "installed". Its just an exe. If you can tell me what the 62/62 should be, I can just switch it to that and save and all will be good. My performance is fine though so maybe I should just leave it.

 

By intel specs the 920XM is 55w. But Dells bios runs it at 62W as that is how dell designed it. But realistically, even though dell limited the CPU to 62W, it will draw about 67W or so.

 

Just and update guys.

I was looking at gpuz this morning and when idling on the desktop i was getting clock spikes from 100/150 to max clocks at random intervals with no load at all.
so i system restored to just after i clean installed the other day and the problem seems to have gone. I've benched F1 2011 and everything SEEMS ok. Not got chance to have a proper look as the mrs has come home so will carry on in the morning. Thanks for the help guys i'll +1 you all when you can. And il keep you informed.

 

Awesome news Clarky...hope all your additional tests go just as well.

 

Any application that uses HTML5 or something similar (even windows aero) can cause that. If I have my computer on idle with a single firefox window open, but I'm not doing anything, it will spike back in forth from 100-150-830mhz and back. As far as I can tell it's firefox b/c when I close firefox, the spikes almost completely go away. I haven't had the time to read up on everything you've posted, but my first inkling is that it's normal.

 

So 62/62 is normal?