M17xR4 issue with GPU 675M / 680M - Help and Advices are welcome

Sticky Summary (edited October 12th 2015):

That thread ended with some good conclusions even if other fixes had to be launched ( see this other thread). I thank the people who answered questions along but would be very grateful if a lot of you can go to the end of this thread (click on Conclusion link below if no time to read ) and answer two questions I left opened.

Core sections of that thread are:

Introduction: When your baby got sick (context of GPU issues) - see below
Chapter 1 : Preparing the patient (= various teardown steps)
Chapter 2 : Walking Dead (= reanimating without mobo fixing)
Chapter 3 : Brand new backbone (= mobo change and its results)
Conclusion

End of sticky section
------------------------------------------------------------------------------------------------

Hello all,

After weeks of fighting I am searching for advices or even fixes. I already browse extensively the threads on related topics from this forum and a bunch of others; so its time for questions and also to share that if it can help somebody in the future.

My computer:
2.5 year old Alienware M17xR4 - Bios A12 - icore7 - GTX 675M (120Hz - 3D LCD) - 24 GB RAM - 240 W
My OS / GPU drivers:
Windows 7-64 - NVIDIA 344.11

I started to experience BSOD and freeze at the begining of August. After verifying I realized the GPU & CPU temperatures were quite hot (85 - 90°C). I did a cleaning and repasting of both CPU and GPU. All went good the 4 weeks after: temp from 50 to 75°C max and no more BSOD.

However early this month the issues came back and with a lot more frequency. After checking all the usual culprits (RAM, drivers, Windows, battery, PSU) I focused on the 675M GPU.
The drivers changes and all the other stuff regarding temperature cooling tricks were tried but failed. So I decided to change the 675M GPU for a 680M.

I received the card 3 days ago and made the upgrade at lightspeed with any problem. I started the computer and it just went ok (130fps on OCCT Power Test during a mere 20 minutes). So I went to bed with a smile.

24 hours later I started the computer to show to a friend the new beast - and then the OCCT test just broke after 4 minutes with a "pink" freeze of the screen !

After that I tried to reboot but no way (it always hung on during the windows 7 loading with a black screen and that's all. Changing drivers and so on didn't fix anything. For sure I was able to run in W7 no-error mode but not anything else.

Yesterday evening after checking the card and repasting it again I tried to boot but forgetting to plug the power: I was then able to boot windows. I then tried OCCT - I know that GPU on battery is weak but I am quite deseperate now - and I reached a crippled 25 fps but with no more freeze after one hour of OCCT test. Then I tried to plug and immediately I got a freeze when trying to open FF. I tested a little bit around this and I realized that if I try to launch Windows on power it fails and if I plug after started on battery it freezes also just after - for example plugging the external monitor is immediately freezing the computer on a screen of various colour but that death is without any BSOD registered and I have to shut down manually. Because I have two 240W PSU I tried both and the issue is the same - so it's not the PSU. I also moved the plug on computer side but not experiencing any short or false contact that way.

I then tried to put back in place my 675M (kept as a souvenir) and I got quite the same conclusion: if I use on battery it works but with power the computer crashes; perhaps after a little bit longer time with the 675M - but a BSOD occurs and is registered by W7 with that GPU. I reproduced this behaviour a lot of time today and I even found that the issues are coming quicker (with the 675M) when the computer is hotter (but far below critical limits).

I also tested all with EPSA just in case but all is detected OK (sic!)

Long introduction but useful to understand my questions :

1 - Do you have an explanation or an experience for that issue ? And why not a last trick I can try?
2 - Should I change the mobo (more $ after the 680M) or buy a new computer ?

Thank you for the feedback folks
Spacetauren

 

Bump! I can't imagine that any of you has even one idea about a root cause.

Anyway, I bought a genuine refurbished mobo and I will go for a full Monty - first step is to receive it from the eBay seller...
After that I expect that I will need some guidance for the next moves to reach an happy end.

First request : Can somebody point me to some useful docs / video or whatever that can help me? I've seen a youtube teardown but not sure it is bullet proof?
Second request: I have a 120Hz 3D model and I would like to know what are the difference to plug the LCD on the mobo. Can somebody explain ?

I think some forum deities like J95 have done almost all that can be done in tuning their M17xR4 and I am still hoping they could help. Is it a way to contact one of them ?

As a contribution I will put in that thread a full report on the next steps of my attempt to raisedead the computer. Followers and contributors for that big show will be welcome.

Come back soon !

 

Ok well, it could be a couple things:
1. Your motherboard has a bad MXM slot that is screwing with your GPUs
2. Your 675M and your 680M are both bad.

I would have said test with the Intel HD 4000 to see if you get similar results, but you have the 120Hz panel so that is not possible to do. As for your requests:
1.
2. I'm not sure what you mean by this. There are 3 panels for the M17x R4. The 1600x900 60Hz and 1920x1080 60Hz panel use the LVDS connector, and the 1920x1080 120Hz panel uses the eDP connector. They look completely different and there is no way you can confuse the two when you replace the system board.

 

Hi and thanks Raidriar.

1. The MXM can be the faulty one but it is only when GPUs are running on PSU and not on battery. What does this imply? Is it possibly linked to some VRMs IC that are no more able to manage the power coming from PSU but OK with battery?
2. This can be true - both GPUs bad - which is a little bit sad for the brand-new 680M I only used 20 minutes with no issue and that failed the day after.

The only way to test more is to change the mobo I'm afraid - or to find another laptop where I can mount these GPUs and test (lucky me!).
Let's say I will first have a very cautious look to the ICs on my GPUs and mobo this week-end and I will revert if I find something strange.

About requests answers: I already had this youtube link but I thank you anyway. I am trying to find more detailed docs but perhaps they are not available on the web.
Greatly appreciate your info on the 120Hz connector. If I cannot be confused then the teardown will be easier.

 

I know that on battery, the GPU will scale back to only 2D clocks and low voltage, impossible to use 3D clocks and 3D voltage, so it may not be straining the card so hard or stressing the board. Not likely to be VRM issue, but either something wrong with the internal GPU die, bad solder joint, or a dead IC or capacitor either on the MXM card or on the motherboard. It would really suck if the 680M is dead, they don't seem to die too often. 675M is a known garbage card that was a rebadge of the GTX 580M, which was notoriously unreliable.

As you said, the only way to know is with a motherboard swap. I believe this is the eDP socket on the board:

 

Happy to read that both of us are sharing the same feeling Raidriar. I really cross my fingers that the 680M didn't died.
What I find quite uneasy to understand is why the 680M were able to do a good job during a stress test at the first time. The idea about a dying capacitor could indeed fit with the behaviour I experienced but with ICs it can be more tricky unfortunately.

Thanks for the picture with the eDP I will check but seems clear.

I will post news about my next moves. I will copy / record all my configuration first and I will after go for a minimum system installation on one spare SSD I have (I will store my other HD & SSD currently mounted so that I can swap them at the end if I fix the mess). And also I have a spare 970M I can use at the very last option (but I will then need to change of OS for 8.1 or 10 I guess).

 

Good luck!

 

I once had it boot to a black screen, I know this is going to sound weird, but hit your Fn+F6 key, then hit enter, it'll select the monitor as your regular monitor.

I don't know why it happens, but it happens only with legacy boot and W7.

You can test your GPU in any laptop that accepts Dell GPUs really. That is a way to guarantee it's the motherboard, but the odds of getting TWO bad GPUs is pretty low, so I will assume it's the motherboard.

 

Thanks danuyne. Nothing is too weird for me these days. I tried the Fn+F6 key after reading your post but doesn't fix my issue . But I will keep this on mind for future adventures with laptops.

OK I will now start the reporting of my raisedead trial.

CHAPTER 1 : Preparing the patient

Before to replace the mobo (still not arrived) I will first remove my SSD and HDD so that I can kill 3 birds with one stone :
1 - Be sure my 2 and a half year data and system are safe during the next teardown steps
2 - Having a fresh OS & drivers install that will avoid (I hope) the occurrence of possible mess of two many softwares and drivers with the new mobo when mounted
3 - No need to do a quite uneasy and hazardous backup / mirror of my system with my unstable computer

Plan is good but then I figured out my SSD is indeed a mSata one. Looking on the web make me realize that this small crap is kept in a M17xR4 like gold in Fort Knox . I will need to do quite a first full teardown to remove it. Ok friends let's consider it like a training before to go for the teardown for the mobo I will have to do anyway.

There we go ! So the very first thing you don't feel so good to see with a laptop is it's bottom as a lot of you I am sure know well.

I will not go for an extra detailed script for this teardown because you have still a lot of sources on the forum for that. But let's have a look of the most accessible components.


After that I have to remove the keyboard and the lcd panel assembly. Which means a good time in disconnecting a bunch of tiny connectors of any kind (quite a creepy time I swear). The below pic give a little feeling of how it can be stressful.


Ok after all of this is done you can remove the under-keyboard layer :


Then you can have a look of the very inside of the m17xR4 (pic below) and realize that this backbone is not as sexy as the computer you had in front of you some minutes ago (and also a lot lighter ).



And finally here is the holy crap mSATA SSD (pic below). It is hard to find but quite easy to unplug .


....
And now it is time for a full reassembly yeahhhh . I will not spend too much time to describe it but let's say it went quite smooth even if I had some butterfingers stages with the keyboard connectors. The mounting of the Samsung EVO 500 Gb SSD in the SATA III first slot (HD0) was a piece of cake.

Next post I will explain the computer and new OS first revival (hopefully) .

 

... and I will need it, thanks raidriar.

Chapter 2 : Walking Dead (= reanimating without mobo fixing)

So after having removed the former mSATA SSD and the HDD and put in place a new SSD on SATA port it's time to check if I didn't kill the computer during the dissassembly / reassembly session.

I have prepared an usb boot key with Windows 7. I turn on the computer and push on F2 to enter the bios. It went fine at that stage and I checked that system seemed ok. I plugged in my usb key and with a shaking finger I exited the BIOS setup. The M17xR4 ran immediately with a long set of bips that I interrupted by pressing the alien head. I panicked a little - is it a miss in the reassembly ?

I had prepared my genuine dvd of Win 64 - always be ready with a plan B is a lesson IT told me a long time ago. I decided to remove the usb and to try a boot with the dvd. I turned on for the second time and … yeeesss !! ... it started to read the stuff and to install W7.

I spent some hours after to update the W7 and to install the basic genuine set of drivers I need - but without installing the one for the nvidia GPU. During that time (almost 10 hours) I had three times a strange black screen - very like a sleep mode but with no way to recover from it (here I tried the Fn+F6 button as per the danuyne post but with no success). However the computer was still working in the backstage because when I stopped and restart with the alien button I had normal restarts of windows with a resume stage in two of the three cases. No clear explanation but could be a mobo issue again. Looks like the message to stop the screen is coming suddenly to the system and that it has no clue on how to come back to live after. No BSOD that times.

At the very end I finally re-installed the original GPU driver that came with the computer 2,5 years ago (rev. 307.17). I started to use the stuff a few minutes (even changing successfully the frequency to 120 Hz) before I had the very well known BSOD with black screen and reboot I have experienced a lot since this story started .

So I am more and more sure that my mobo is not in good shape, but I am not yet sure that the two GPUs have not themselves been damaged. Now I need just to wait for the mobo and pray it will be a fine and robust one.

Any new inputs, advices and encouragements of people that made the same trip - mobo replacement by themselves and so on - are more than welcome.
Next chapter soon (I hope)

 

Hey. I had two extra m17x r4 mobo: you should have sent me a pm first.
I would recommend trying the gtx 970m since normally the mobo wont go bad easily.
I bet it should be your gpu's.
Before you try the gtx 970: you should change some settings in bios.
That has to be done with gtx 680 or gtx 675 because if you try to boot with gtx 970m it would throw up a blank screen.
You have to change this settings in boot menu: boot list option: UEFI , load legacy option ROM: disable , secure boot: disable , fast boot: disable.
Now save and exit.
Turn off the lappy and install the gtx 970m .
Make sure you have a usb drive with windows 8.1 ready.
Install the OS and so the drivers.
Run some benchies and see if you still get the BSOD's.

 

Gasp @nbr2019 ! If you are right I can say bye bye to my 680M at 800 bucks (with all taxes and shipping costs - I'm in France).

Thanks for the advices & method to install the 970M. But I am hesitating because I am afraid the mobo can damage the GPUs I plug on it. If you recall what I wrote above in one of the first post: my 680M had a good start and it is only on the second day that things became odd. Here comes my bad feeling: "Does the mobo kills the GPU I plug ?"

Note: for the spare mobo you have, I couldn't know - and also I tried to find one close to me (in UK indeed). But I keep the information just in case I needed a second replacement mobo

 

800 bucks no way ? Is it GBP ?
I would buy two gtx 970m for that price.

I don't think mobo can ever kill a gpu.
Let me know how it goes with new mobo when you receive it.

 

Chapter 3 : Brand new backbone (= mobo change and its results)


I received the new mobo and went to replace the old one by this one.
I had still completed before (see above in the thread) most of the full teardown when removing the mSATA SSD.
I went through all the steps of the youtube video (see first reply of Raidriar on page 1 of this thread). At the end it was missing few indications to actually remove the mobo. However you can find them here.

In order to make the job easy you need to organize a little bit - I chose to use three large white paper sheets in order to put in a kind of sequence the various components and screws I disassembled.

NOTE: don't forget to get the few tiny components that are still on the mobo to put them back on the new one (should be an mSATA SDD, a bluetooth card, some RAM, ...).

I picked up the new mobo and I did reverse sequence for reassembling the computer. Quite easy with my white sheets trick. I put in placed the 680M (the new costly one that worked 20 minutes when I did the first upgrade of the 675M - see the story in prior posts).

Ok I plug power and try the first boot : beeeeeepppp. I shutdown and boot again … and Windows 7 is booting but hangs on a black screen. Damned it is the same sh.. than before. I power off with the alienhead and I unplug the power. Now I boot on battery, and Windows 7 launch correctly. I power plug then and few seconds after I get a white screen. More attempts with alternate drivers just had the exact same effects.

After that I removed the GPU and put the old one 675M with a proven driver I used during years. I did that step just to confirm the new mobo is fully operational. After a lot of tests - but not stressful for the GPU - I can conclude the new mobo (A11 BIOS) is fine. By the way I even recovered one feature I had no more for ages - if I ever had it - on the older one: the ability to run computer when battery is removed.

At that stage I launched a stress test on the GPU that failed after 8 minutes with a BSOD.
Same as before. No surprise

 

Conclusion: some learnings and a few remaining questions

- 100% sure : brand new 680M is dead !
- 100% sure : old 675M is crippled ! (but not fully dead)
- 100% sure : new mobo at least as fine than the old one (with a little plus)
- 95% sure : my old mobo was likely not the root cause of the first issue
- 100% sure : It's OK for a cautious one to sucessfully teardown + reassemble a M17xR4 with not so pain.

I hope this one thread can be useful for other M17xR4 / R3 users in the future.
I will continue in another thread for next moves (do you recall my spare 970M?).

Still have two questions:

1. Do you think I can be refund (or at least get a new replacement GPU 680M) from the retailer - Eurocom - that supplied it to me for more that 800 USD ?
2. I am in Europe (France), do some of you had a positive experience about such a kind of replacement of faulty GPU with Eurocom ?

 

-

 

As I said earlier, it was the gpu's .
You should have taken my word on that.
Yes eurocom offers warranty. Send them a email and get it sorted out.

 

Thanks nbr2019 I will contact Eurocom. Indeed I trusted you but the mobo was already on its way. Was then a good experience to replace it anyway.
I am now installing and tuning a 970M + W10... works but not easy and perhaps not stable.