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

Nice mate Well done!!

As per the driver discussion, I am still using 12.4's......any benefit to try newer pre-release ones?

 

Anyone got a link to the dual 240 PSU setup?

 

http://forum.notebookreview.com/alienware-m17x/585064-guide-installing-amd-radeon-mobility-6970-6990-your-m17x-r2-single-gpu-crossfirex-95.html

It's earlier in this topic. Start reading from page ~70-95

 

So, I'm having the SAME problem again with my laptop. Just fired up my laptop for the third time with the 6900's xFired. I noticed that I did get one of the cards back, but I got one new one when they returned them after testing. I am getting this information which looks WAY wrong from HWiNFO64. I can't figure out for the life of me what could be wrong except the hardware. This laptop was rocking two 5870 cards previously just fine for months. Here is the information: http://i857.photobucket.com/albums/ab134/soccermadman/Information.png

So laptop showed normal after a full cleaning and reinstall in HWiNFO. It still freezes after about 60 seconds of having Diablo3 on the screen. I am ready to email Eurocom and ask to send the 1 card they sent back to me and send it back for replacement.

So here is some more interesting information - Check the information in the screenshot. I am THOROUGHLY confused:


I can't figure out why the default clocks look soooooo high, and I don't know why this was stable for hours the other night and now, once again it will NOT work. Its OK on the windows desktop this time, but I can't for the life of me figure out what is going on.

Douse, I've been reading through the forums, and read a number of posts about the 6990's you've posted in some of the threads, and I know you are using 6990's. I see that the cards that I have are showing as manufactured by Dell (according to GPU-Z). Should I take the blackcomb.bin that you uploaded in another thread and flash that on these cards? I'm at my whit's end and I'm actually pissed enough about this problem (I HATE leaving a problem unsolved) that i've stayed up til almost 2 AM now on less than 5 hours sleep for the past 2 days total. That said, I'm either retarded, I've missed and repeatedly overlooked something simple and I'm going to feel like an when I figure it out, or I'm losing my marbles. And none of this is anyone's fault here, but I'm hoping someone here has an idea.

Edit: I guess I have to cave, falling asleep in the chair here and my wife (who's had LOTS of sleep in the past few days already caved and hit the hay). TIA for any pointers here.

 

The freezing is the Machine_Check_Exception BSOD.

Essentially what's happening is that the GPUs are drawing too much power from the PSU and the CPU is physically stopping the machine.

I had this problem with my 6970Ms.

I used Zfactor's guide to creating a bootable DOS disk and flashed both using a modified vBIOS, changed in the program RBE, and using ATIFlash.

-Notes-

-Use the updated guide here and use his win98boot file for the flash. It's easier to work with.
-Use the program called 'RBE' to modify the vBIOS. (PM me if you want help with this!)
-Use ATIFlash when you do flash the GPUs. Much safer and super easy to work with.
-Make sure to use HW Monitor to check if the voltages changed after the flash.
-Do not overclock anything, leave it all at stock; even the CPU.

 

So, what did you change in RBE? Am I undervolting the cards? I've removed the OC on the CPU but same result. Should I buy a second 240w and make a 480w PSU? THANK YOU FOR YOUR HELP! I was having a bit of a meltdown here lol

 

You definitely could try undervolting the cards if you have not already. I think most folks are fairly stable at 1.0...I'm running .96 on my 6970s and have been very stable, so worth trying to undervolt it a bit and see if it helps.

Also make sure you are running stock on your clocks (no overclock in the bios, no overclock of the processor or the GPU's) and see if that helps.

Just a few things to try...you can use Trixx to undervolt the cards before using RBE to set them and see if that helps. If you don't have a Kill-a-watt meter you could purchase one of those as well (Home Depot, Fry's electronic's etc) which will tell you if you are going over the 240w when you are playing Diablo 3 (or any other games)

Just some things to try.

 

I really appreciate the information Grey. When I even try to start 3D stuff with Crossfire enabled and both cards in, it freezes INSTANTLY now. That said, I managed to play 2 hours of Diablo 3 the other night, so I don't really get it. The thermals were fine the whole time. I'm kind of wondering if the 1 card I was sent back from the first batch was the bad one and not the good one. I'm going to try running them single card at a time tonight (swap the one i think might be bad into the main slot and run it by itself).

I'm really frustrated at this point, cause I can't even necessarily stay stable at the desktop once again tonight. I've attached a CSV from HWinfo from my last D3 launch, some interesting very weird utilization information in there.

I removed the OC on the BIOS, but I'm wondering if I need to reflash my modified bios as well. After a lockup my laptop won't boot unless i leave it off for a couple minutes. It just sits there blinking the scroll-lock and caps-lock while num-lock stays solid.

HALP! lol

Edit: I can't lower the voltage on the secondary card in Trixx. It won't let me. Is that a bad sign?

Edit:2: It now shows (earlier this evening the cards showed as Clevo/SOMETHING cards) as Dell cards again. is up with this thing?

 

Hey PWn3R,

Man, sorry for all the problems you are having...I completely understand your frustration though (went through my fair share of it in the beginning of this thread...hard to believe that was a year ago). Anyway a couple more thoughts and forgive me if I'm repeating stuff you've already done..it's late so I'm not thinking 100%

First of all, trying one card at a time is a great idea. The games should play (albeit slower) in a single card configuration. If either card freeze while you are playing then it's highly likely one of the cards you have is still bad.

So I think that's a great idea. Also, re-flashing the vbios on the cards isn't a bad idea and worth a shot (done mine several times and so far so good). Just to make sure that the vbios isn't somehow causing the problem

Regarding Trixx and not seeing the voltage on the secondary card. That might very well be a Trixx issue and not an issue with your card. I recall doing a search where there was a change that had to be made for Trixx to be able to modify the voltage on the secondary card. Do a quick google on it and see if something comes up (I'll paste it here if I come across it as well). But I'm fairly cetain that's a Trixx problem, so don't be TOO concerned about that specifically.

*EDIT* Found this, it's older but may still apply http://www.tomshardware.com/forum/274898-29-sapphire-trixx-unable-view-change-clocks-before

I'll admit that I flash mine using RBE these days so haven't messed with trixx since last year.

Regarding the cards showing as Clevo before and now Dell...was that after you flashed the vbios again? Did you use the correct vbios? Just some thoughts have no clue what might have caused that.

So now that I've rambled a bit...

I'd say disable crossfire and try to run on a single card...if it runs fine, then swap the secondary card to the primary and repeat. If both run fine as single cards then it could be a crossfire cable issue, power issue, or something else. But at least narrow it down first that way.

Keep us posted on how it goes..and hang in there...trust me, took me months to get mine working (long story but it's all documented somewhere in this thread in all my frustrated glory)

Good luck!

 

Greywolf, Thanks for the information. I am in the process of tearing down the secondary card to mount it on the primary heatsink to see if it is stable. I don't have a stock known good Clevo BIOS, and the version that came on these cards is 1000's of rev's higher than what I can find anywhere. The cards should be Clevo cards as I bought them from Eurocom. I also have no idea how they showed Clevo and then Dell (i've seen that a couple times now). I didn't flash them or do anything except get a lock up on the machine. I was able to launch D3 without issue with Xfire off.

Did anyone have to mess with the settings in the modified BIOS to make everything happy?

Edit: I'm really hoping the secondary card is bad. Otherwise, I'm not sure what to do. This thing worked fine with the 5870's.

Edit:2: So, I've found that the second card goes to a black screen and locksup within seconds of booting into windows. This is also the card I suspected was bad the last time around (i took pictures and the label serial # is the same. I am going to contact the guys at the Eurocom store and find out what they want to do. This is highly irritating - I suspected this was probably the problem, I'm new to modding laptops but I'm no tech dummy, by day I'm a Citrix/Windows/AD Admin for a hospital system, and the team lead for the Wintel group.

 

Hi PWn3R,

It does sound like that card is indeed the problem. Hopefully they'll just exchange it for another and you can get it up and running. Now that they claim support for the alienware system it shouldn't be an issue getting a replacement...but then I've never had to deal with them on that so good luck to you.

 

Was there ever a VBIOS release for the 6990m to enable sound over HDMI/DP?

 

@PWn3r, get your card replaced. Same thing happened to me...2 times over. The 3rd card worked, until I killed it. The one thing I did differently with the 3rd card is I didn't tighten down the heatsink as much as I did the previous 2. Not sure if this was the reason it worked or I just finally got a working card the 3rd time around. Hopefully they will replace it. I think there were a lot of bad eggs in the 6970/6990 batches.

 

may i ask you guys what type of thermal pads did you use? i just bought 2 x 6970m's from a fellow member and i am using my 4870 heatsinks for the mod. i would like to know what thermal pads to buy off eBay to have them here ready for the installation? i mean size and thickness.

cheers
tio

 

I used .5mm from frozencpu.com

 

do you have a link mate?

 

http://www.frozencpu.com/cat/l2/g8/c487/list/p1/Thermal-Thermal_Pads_Tape.html Brand doesn't make much differance, I've used the EK Systems and the Fujipoly.

 

Like DR650, I also use the Fujipoly in mine...

.5mm currently, although some folks say to use the 1mm. I'd definitely purchase the .5mm and you can always stack them if they aren't making good contact.

 

yes guys i understand what you mean but,in the link i dont see anywhere 20x20mm

 

Hey Radji,

No VBIOS was released to allow sound over HDMI...I believe a few pages back it was mentioned that sound works of DP (unless I'm getting that confused with something else).

Regardless Sound over HDMI doesn't work unless you are using Dell branded cards.

Hey TB303,

Try this link:

Fujipoly Premium System Builder Thermal Pad - 1/4 Sheet - 150 x 100 x 0.5 - Thermal Conductivity 6.0 W/mK - FrozenCPU.com

The 1/4 sheet should give you plenty to do both cards and double them up if you need too.

 

Thanks! so i get to cut my own?
what about this here: High grade thermal material
anyone figure it out?

 

Damn. The Dell cards still don't have fan control & crash on shut down still, yes?

 

I wouldn't go any thicker than .5mm thick. I used the Fujipoly stuff myself and the GPU isn't overheating in any regard...so long as I don't disable Vsync.

Well... yes, I'm using two Dell branded 6970Ms at the moment but since my vBIOS flash, they'll successfully shut down when I shut down. They just don't like it when I unplug the power cord...and they don't go to sleep. It's either on or off.

But no they'll work fine. *points to sig* I use HWiNFO64 to control the fans. Here's the profile setup I use for the fans and CPU. (images attached)

(*points to Network ID Name in picture* Courtesy of my mom. )

 

@tb303 Yeah cut yourself...I used the old ones as a guide and then just trimmed them as needed. Not hard at all, and you'll get a lot of pads from the 1 sheet (I even included some pre-cut ones when I sold my other set of 6970's). Can't speak to the high grade thermal material that mxm sells, but I imagine it's all relatively the same. I will say the fujiopoly have kept my temps nice and low, been very happy with it, but I imagine most brands are similiar.

@radji Correct about the dell cards, you can control the fans using HWinfo32/64 but that is still a requirement as is the shutdown issues.

 

WOW. That sounded so bad... sorry... had to comment... XD

 

hahaha...ok so maybe I should have said "cut the pads to fit...yourself".

On second thought I don't think there is a good way to say that, so I'll just leave it alone

 

i am using those as well.

 

would this work?
Arctic Silver Thermal Adhesive

 

Some people have found that .5mm doesn't make the proper connection with the memory chips on the card. I started off using .5mm and then switched to 1mm after noticing this myself. No heat issues with 1mm pads but better contact for me.

Fugipoly from FrozenCPU is what I used.

edit:

That would not be recommended. It is not for use with CPUs so it's safe to say that it shouldn't be used on GPUs either. You don't want anything permanent attaching the GPU to the heatsink.

 

Which revision of the Clevo vbios are you using? I used to run a dell card. On the earliest revision of the Clevo vbios, my cards behaved exactly as you described.

However on the 2nd revision, my cards were free of ALL problems except for the fans.

 

For my M17x R2 in my signature, how's this 3DMark11 score look?

AMD Radeon HD 6970M / 6990M video card benchmark result - Intel Core i7-950 Processor,Alienware score: P5385 3DMarks

GPUs are undervolted to 1v and slightly underclocked to 650/900 vs 680/900.

I mean... according to Notebookcheck.net, the 6970Ms in Crossfire only hit 5541 with a i7-990X...yet my undervolted GPUs are nearly hitting 5700 for the GPU score??

Are these actually running more efficiently AFTER being modified?

 

what is the best thermal compound i could use on the CPU and the GPU's guys?
I am buying the Phobya thermal pad sheet 100 x 100 x 1.0 mm for the Vram as it is the only one i can find in the UK which makes delivery a lot faster as well as that i will avoid taxes. The Phobya is the white one that has a thermal conductivity of 7W/mK compared to 6W/mK from the Fujipoly.
Now that sheet will be good for 25 pads therefore it is quite enough as long as it is good. Anyone using it?

 

Well...the best thermal compound you can get is the Indigo Xtreme, but that's desktop exclusive. For laptops, the best is a toss-up between IC7 and Shin Etsu X23-7783D.

Both of these are incredibly expensive though. I got my last 1.5g tube of the X23-7783D stuff for about $8. It was worth it though. Temps dropped for a solid 7*C all around. However, the best runner up for these is GC-Extreme. It's crazy cheap by comparison and I can't tell a difference between the Shin Etsu and the GC-Extreme thermal material as far as temperatures go. I repasted my GPUs with the GC-Extreme and temps dropped a solid 10*C vs the AS5 stuff I used before. Same coverage both ways.

As for thermal pads, I used the Fujipoly 0.5mm set and my temps aren't at all high...all things considered of course... It's perfect coverage as far as I can see.

 

i guess i screwed it up then huh?
this is what i bought:
Phobya XT 100 x 100 x 1mm Self Adhesive Thermal Pad
Gelid GC-Extreme Cutting Edge Thermal Compound 3.5g
Arctic Cooling MX-4 Thermal Compound 4g

 

Greywolf, they messaged this morning (2 days after my last message on ebay) and said they wanted my email address to talk to me about the issue. No other updates today. I'm getting the feeling they are sick of me. If I have to, I will open a dispute with eBay, but I'll give them til tomorrow evening to response. They said their support would contact me directly. I can run on one card for the time being, but I put off repasting properly after the testing with the card that I know is bad now, so my temps have been running at about 85C after 2 hours at load. I was hoping to wait til I had to tear it down again to put the replacement card in there. Hopefully I can get them to replace the card. They swear they tested both of them twice before packaging and mailing them to me, but I don't know how you could've tested the card and when I put it in it sort of works, unless all then did was throw it in and boot to windows. I was able to do that a few times, but its never stable. I'm hesitant to flash the card with a copy of the bios from the working card especially if they offer to RMA and i've totally bricked it.

 

I wouldn't try to use the 1.0 padding with the cards. Atleast from what i've seen, 1.0mm would be way too tall and you would end up with like .5mm of thermal compound between the GPU and heatsink.

 

Actually 1mm should work..I've used both and I saw good results with either .5mm or 1mm...depends on how much it compresses but you may be ok.

Arctic Cooling MX-4 I've heard should work just fine (correct me if I'm wrong, I haven't used it)...I use OCZ Freeze myself which is relatively inexpensive as paste goes and works incredibly well (again my opinnion only, other will vary).

 

Hi PWn3R,

They should be getting in contact with you, but I have heard they can be a bit slow on the response. Have you tried calling their support?

Eurocom Corporation - Number 1 in Desktop Replacement Notebook Technology

I haven't personally tried this route, but they may be able to help as well.

I know some folks have had to do more then one replacement and it has worked out for them, so I hope they are willing to work with you on this. Definitely sounds like the card is the problem, and should be eligible for RMA IMHO.

Anyway, good luck and keep us posted.

 

so what is the problem if it does get compressed to 0.5mm in the end?

 

At last 12.6 mobility officially out

 

1.0mm pads are fine and provide better contact to the end RAM modules because the heatsink normally doesn't fit evenly on the back bank of RAM and the GPU. There will probably be a small gap with the .5mm pads in this area (there was for me). I've found that if you use the 1.0mm pads you don't have to worry about this. Temps are still low with the thicker pads. And the stock pads were at least 1mm thick so it is fine.

As far as the compound goes, I used IC Diamond 7. Hard to spread but cools very nice.

 

hi there!
i got two 6770 cards. is any of the two the mastercard or they are both the same?

 

Ok so I took 2 240w power supplies soldered black to black green to green and white to white and so it works like this >>>------(Computer) where the >>> is the 2 power supplies and the ---- is the single cable I spliced them into and (computer) is obv the alienware. But I no has power what am I doing wrong?

Sent from my Transformer Prime TF300T using Tapatalk 2

 

I need to buy new thermopads for gpu cooler (5870m)
0.5mm is ok?

 

You have a poor connection or something is shorting out.

Additionally, that set up will not work. One PSU, the one with the higher output voltage will be biased over the other resulting in basically no current being pulled from the other PSU. You need to isolate them with a pair of schottky diodes housed in the same IC. That is the only way it will work.

 

hello,

i installed the 2 x 6970m cards and managed to get crossfire. how do i check the temps please? i use to have rivatuner for that but now no luck. i used msi afterburner which show 47 degres on gpu 1 and 67 on gpu 2. then i removed the gpu 2 which i thought was probably the slave card to the right of the mobo, in order to check the thermal paste and the pads. then i reinstalled and i cannot use msi afterburner anymore. it just freezes my pc. also the same happens in hwinfo32 whenever i try to select the second card. any ideas please? is the card overheating or something?

 

check this out.

this is on idle.

GPU0


GPU1


Now with FurMark (multi-GPU) running - under load.

GPU0


GPU1


what is the second GPU doing?
Why is it clocked that high on idle?
Also under load why does core clock stay so low?

any ideas please?

 

It has to be fullscreen to fully use both GPUs. It was doing that in 3DMark11 too when I used it in windowed mode. I would also HIGHLY recommend that you undervolt the GPUs to at LEAST 1.0v if not a little lower. Your machine will probably seize up and BSOD if you run both at maximum power. I still freeze from time to time if I don't have Vsync on in game.



Also, to the guy who was asking about the thermal pads, 0.5mm pads are what I used on my 6970Ms and these seem to be just find for the job.

 

I've been digging through thus thread for the last hour and haven't seen any info regarding this. Is there a step-by-step somewhere?

Sent from my Transformer Prime TF300T using Tapatalk 2

 

Read from here - pg 70 - through page 100 or so. You will see how the setup you posted doesn't work and what exactly has to be added to make it functional.