P170SM-A from a (probably) dying GTX 880M to a 980M Upgrade

Dear all,

I have a (probably) dying GTX 880M from my P170SM-A that I bought from PC Specialist back in 2014. I say probably because I am assuming that is dying due to the error of the card being constantly recognized as a "GTX 780M" in the BIOS generating constant Blue Screens of Death. In addition, when I disable the card in device manager, Windows runs fine. I made some research on the web and have found that some Kepler cards have part of its ID on the resistor, and it is not unsual to have, for instance, cards being identified as Nvidias Quadros due to the same problem.

Given this fact, I started to do some research on the web I have found this forum and its amazing community with people trying that help each other. I have found some people, last year, asking the same thing of the title of this thread. ( http://forum.notebookreview.com/thr...a-sager-np8278-gtx-870m.831368/#post-10975586). I am, just like this guy, a novice and not too skilled.

However, before I found the actual reason of my hardware problem (appears to be the GTX 880M,since is acting crazy going to a 780M randomly on the BIOS), I updated my BIOS using one of the threads of this forum and all went smooth, I was trying to rule out the possibilities of the given problem ( http://forum.notebookreview.com/thr...-released-in-after-2014.790474/#post-10238079). I used the most recent BIOS available for my card from the thread (may 2015, BIOS EC V 1.03.09
& BIOS V 1.03.11).

Given this, I am decided to upgrade to a 980m. It will be plug and play?

I am sorry for the long post, I also would like to share this interesting thing regarding the resistor of the GTX 880m making it identify itself as a 780m

Kind Regards.

edit: just grammar

 

Ah, a fellow P170SM-A owner !

Fwiw, I just upgraded to my 980M yesterday (my 880M died too, not in the same way though), but didn't check Windows yet. However I checked on Linux and played The Signal from Tölva for some time without any problems.

The installation was rather plug-n-play, except for that metal casing attached to the 880M, whose screws didn't fit any more in the 980M, so I put in back inside without those screws but it seems fastened enough.

If you didn't upgrade yet, maybe you should benefit, in addition to thermal paste of course, from buying some thermal padding of two thickness sizes (which I forgot about, hence I had to do with some original thermal padding, that was slightly dirty and/or damaged).

In case you upgraded already, I would be interested in knowing how it fares for you with the newest BIOS/EC, since I didn't flash yet. Most notably, does the 980M get recognised by your BIOS ? (mine says "GPU : unknown")

 

Yeah the back plate is not needed if that's what you meant.

 

Then I wonder why it was present in the first place... I mean, it's not a small thing after all, and removing it from the manufacturing process surely would have saved a nice sum to Clevo. That being said, I don't feel like removing it, now that everything is in place, unless of course it can pose problems.

Speaking of problems : I finally updated my EC+BIOS, and indeed, my 980M is now recognised, in the setup screen. However, I have experienced two very problematic freezes on Windows, the first one while updating the Intel HD Graphics driver for my CPU to the latest (when the screen blinks, at some point it went off and remained so + keyboard unresponsive), and a second time while logging in as some user.

Those two freezes were not temporary, and after 20 to 30 minutes I decided to force shutdown (long press on the power button). Previously this was ultra rare, maybe what, once every two years...

It seems to correpond to times when the dGPU is solicited, you know, those times when despite not using it at all, MS or Nvidia (I don't know which one for sure) decides it has to switch on anyway, such as when logging in/out, launching/quitting the task manager and some other mysterious situations. Because I was not using it when Windows froze, but the hybrid graphics light was on anyway (and remained so during the freeze).

Ah ! And well I just got a blue screen now, the error being : DPC_WATCHDOG_VIOLATION. I will have to investigate on this one, but I will probably do that on Linux, as I feel I will have far less problems there...

Just in case, has anyone tried a BIOS + EC downgrade already ?

 

Hello,

Hope my experience can help you, i own a P150SM with a 980M, if i'm not wrong it's the same laptop, just smaller since bios updates are in the same group.
So i discover tha using latest intel VGA drivers made me unable to lunch Forza Horizon 3, BUT lunching 3D Mark Time Spy made the sceen to display fine. Forza Horizon 3 is the only game that i discover not working for me, and the 4th is strangely working fine... Symptoms are app close just after the Microsoft screenwhile loading. I discover that disabling intel HD 4600 made the game to run, at a very low FPS.
I tried, after uninstalling the latest Intel HD drivers, using Windows Update to install "older" drivers and using ones from 2017 made the game to lunch! BUT Time Spy strangely display as if it's stuttering, although the score is good (around 3150pts).

So here what i understand, i hope others can correct me if i'm wrong :

Intel HD 4600 in our laptops are just here for two purposes : as a GPU for low power requirement, and, more usefull, as a display interface for the 980m. I think it's why we can't totally disable the iGPU and use only the 980m, ther is no display directly connected to it.

If you experience other freezes again, maybe give it a try : use older drivers for your HD4600 and always the latest for your 980m

BTW, i'm french, i'm new here, but really not new on computing

EDIT : forgot to answer your 2nd answer : i downgraded from a P150SM-A 1.03.11 to a bios from P150SM 1.03.05. Too bad i did for nothing, and i bought the computer with Prema bios installed, but without a copy. I miss the memory settings...
So I think you can downgrade, just take to flash BIOS and EC together. I don't think it will solve your problem. Hope you can install linux, a lot of settings are hidden in bios and some need to be activated to install linux properly. Try a live USB first

 

Je suis french moi aussi !

Concerning your symptoms, you're lucky you only had returns to the desktop, freezes or BSODs are the very least I can get.

In fact considering I had those problems without even explicitly using the dGPU, I didn't dare launching a 3D app. Additionally, I tried launching the control panel, which resulted in an immediate freeze too.

Right now I am on Linux where it's stable now simply because I'm not using the dGPU (Nvidia Prime Intel profile selected). As on Windows however, each try at my 980M since the BIOS+KBC/EC upgrade resulted in a freeze. Previous to the BIOS upgrade I was able to test with a 3D intensive game successfully, as I mentioned in my previous post.

The interesting thing on Linux is that I've some Alt+SysRq key combinations activated, but even these didn't get me out of a freeze.

This tends to confirm the impression I had on Windows that almost if not all code execution is interrupted, all IRQs are blocked, no more network, no more keyboard, no more nothing.

Moreover, even just calling control methods for the dGPU such as _ON or _OFF could yield to a freeze, even if the Nvidia driver isn't loaded and the PCI device not registered in the kernel. This added to ACPI AE_AML_LOOP_TIMEOUT errors while calling said methods, really makes me think there's a serious problem with the BIOS.

I wonder why it too often seems BIOS developers are severely incompetent. I don't think I experienced once in my whole life a BIOS update that went smoothly... and I'm only speaking of official updates here.

Concerning Windows however, I tend to confirm your diagnosis : I managed to do a bunch of updates (had not been running Windows for two months), all this with the BIOS already fully upgraded and the latest Nvidia driver installed, without a single crash. The first and subsequent freezes/BSODs only occured during and after the Intel Graphics driver update, which I installed last.

Now I'm entertaining the idea that maybe, uninstalling/reinstalling the Nvidia driver after the latest Intel driver would solve the problem. Or maybe it used to work because the previous intel driver was installed prior to the BIOS upgrade...

Finally, I bought my laptop at LDLC, who had their logo displayed during boot (wiped during the BIOS upgrade). That means they injected their logo into the BIOS, maybe they did other modifications I'm not aware of ? While I doubt they did that on their own, maybe they could have used, say, a Sager BIOS ? I don't know actually.

 

The arcane arts of bios programming

 

Hello, Fellow P170SM-A Owner! It is nice to nice another person still running these beasts, besides me

Regarding the upgrade, I bought a GTX 980M at aliexpress and it has arrived my home without issues. I have installed it and checked in GPU-Z and it was a genuine one. I am using it for a month and even playing recent heavy game such as Red Dead Redemption 2, Mafia: Definitive Edition, DOOM Ethernal and so on, All of them at high @ 1080P (except RDR2 on medium, but it is fine). The GTX 980M it is a BEAST. It breath new life to my P170SM-A.

Regarding the BIOS, I went to a post that exists in this forum, a repository for the CLEVO Laptops BIOS/EC, And downloaded the most recent one for our rigs (IIRC, from march 2015 I guess). Then, I updated the BIOS following the instructions from the text file (It is a one-step procedure: From my research, this was made by PREMA, he created a single file to execute the BIOS and the EC in the correct order to do not have any issues) and it was fine. When I plugged in the GTX 980M, it was recognized properly.

Note: I used the most recent STOCK bios for our P170SM-A from the repo from this forum, and I updated the BIOS before installing the gtx 980M.

 

In addition for in folks, I have not experienced any crashes, BSODs, or anything related to that.

My P170SM-A right now is running EC V 1.03.09 & BIOS V 1.03.11, and I have downloaded the files from http://forum.notebookreview.com/thr...-released-in-after-2014.790474/#post-10238079

No problems at all. All went fine in both updating the BIOS and installing the new GTX 980M. I have replaced all thermal pads and used the backplate from the GTX 880M which fitted perfectly (because of the MXM Format.)

While playing games, I have around 75-80°C , and about 35-37°C idle. The only exception is Red Dead Redemption 2 that takes the GPU to up 86°C (Does anyone know if this is a safe temp?).

 

@victor_alves It's good news that it works for you ! That means I have some hope, without needing to downgrade.

I would like to rule out the idea that we have significant hardware or software differences, please tell me if you notice anything different on your side.

LAPTOP : I say my laptop is a P170SM-A because that's what is written on the chassis as the Product Code. P170SM is written as the Model. Is it the same for you ?

dGPU : Concerning the GPU, mine is showing as a GTX 980M, but there still can be differences. I bought it on Aliexpress too, in the bag containing the GPU there was the mention "GTX 980M MSI". If I ever manage to do this I will try running GPU-Z or its venerable ancestor Nvidia Inspector, which I think essentially gives the same info. When you say it is genuine, you say that based on the GPU-Z info only or is there anything else ?

BIOS : I saw the same forum post as you about updating the BIOS, however instead of using the provided mirror, I managed to get these from Clevo directly, despite the MS Azure cloud system they switched to being more difficult to browse than their original FTP system. Here are the links I used :

• EC/KBC : https://enstrong.blob.core.windows.net/en-driver/ALLBIOS/P1xxSM-A/P17xSMA_E0309.zip
• BIOS : https://enstrong.blob.core.windows.net/en-driver/ALLBIOS/P1xxSM-A/P17xSMA_B0311.zip

As mentioned in the article, I updated the EC/KBC first, then the BIOS. Contrary to you, I installed the GPU in the laptop before doing the EC+BIOS flashing, but I only booted Windows (and installed the driver) after the flashing (the old 880 driver had been removed before removing the old card). Apparently the flashing went well, without any error. My original versions were KBC/EC 1.03.02 and BIOS 1.03.03. The GPU Performance Scaling option is disabled in the BIOS since I don't plan to do any overclocking.

OS : I use Windows 10 Pro 64-bit, v1903.

Drivers : I installed the latest Intel driver, v15.40.46.5144, available here. For Nvidia, I installed the latest GRD WHQL driver for the GTX 980M, v456.71 that is. The Intel driver was installed last, and that's when I experienced my first problem. Maybe I should have installed it first, then the Nvidia driver. I will try removing/reinstalling anew the Nvidia driver in case it would help. Alternatively, I may revert to an older iGPU driver in case that's the problem.

Concerning your operating temperatures, I don't know for sure about the 980M, but for my 880M the default throttling temperature was 93°C, and the dGPU fan was reaching max speed at 89°C. I was considering temps below the 90s as acceptable, but generally was feeling more comfortable staying below 80 in the long run. I sometimes accepted temps going as high as 93-94 while playing, but only if it's not for too long and if I can't make satisfying compromises in the game performance settings. The throttling ocurring at 93, reaching 95 was at most extremely rare, in fact I'm not even sure I saw it once.

You mentioned the backplate fitting perfectly, but what about the four screws fastening the backplate to the GPU ? They didn't fit any more in my 980M, so I had to put the backplate without the screws, since I had only four new screws provided. however I considered the presence of screws 5, 6 and 7 (bottom, memory part of the heatsink), plus the two bigger ones (fastening the GPU to the chassis), to be enough to keep it in place.

 

Salut!

Remember, P170SM BIOS may work on your computer, i did flash my P150SM with a P150SM-A BIOS and it worked, still my KB backlight was not working. I don't know exactlly what the difference is here. And that's why i try to help you there, i'm pretty sure P150SM, P170SM and the A variants use the same hardware.

My 980M come from Aliexpress too and is a CLEVO/KAPOK variant with a green PCB like this one. This is NOT the G-SYNC one, GPU-Z show the ID as 10DE-13D7 and confirm the vendor is CLEVO.
That make me think this may be the problem : some weeks ago i read a thread on the web saying how to change the ID of the card using an HEX editor for MSI laptop. Basically searching at a specific adress, should see the card ID and change them to something that match the NVIDIA driver so you don't have to edit the .inf anymore.
I did try, but never found the ID of the 980M at the specified adress BUT the ID of the intel HD4600 instead, and believe me, i tried on some other bioses found on the web (techpowerup have lots of one on the unverified channel) and all of the CLEVO/DELL/Alienware i tested showed the same.

About your MB BIOS this seems to be the latest, i don't think the problem is from here.
About your Intel driver, in my Device Manager, the driver is 20.19.15.5063 from the 29/10/2018, installed with the optionnal opdates from Windows Update.

Your OS is out of date, you may try to update to the 20H2 using the Microsoft Tool. If you have problem, you can still revert back to your v1903.

Hope this help, but i begin to think it's an incompatibility, i do hope i'm wrong.

HS : Do you have a french KB? If so, is the FN and Win key inverted? Just to know, i opened a thread for this since it don't work on mine.

 

Hello Friend! I will try to be as detailed as possible, I know how upset one can be when it is trying to solve a problem like this one and has little information to work with. There it goes:

LAPTOP: The same as yours. P170SM-A, With the same exact chassis and Model. The "-A" is just a revision that was done in our model, IIRC from my readings at the web. So, we are good on this point.

dGPU: My basic inspection to conclude that it was a genuine GTX 980M was taking a look at the GPU-Z main screen and seeing that the clocks, hardware ID and so on, were the ones that are supposed to be for a GTX 980M:
https://imgur.com/a/brvMZsd

BIOS: I am running EC V 1.03.09 & BIOS V 1.03.11. The reason why I used the mirror from the post is pretty simple. I saw that for the files for the mirror, PREMA just got the STOCK BIOS and created a single *.bat file that makes a one-step procedure to update both the BIOS and EC. As I am an unexperienced in the "art" of BIOS Updating , I just assumed that using the files that were an one-step procedure would minimize the possibility of me ruining my Laptop. And I was correct: Thanks to the files made by PREMA, the process was flawless.

OS: Using Windows 10 home 64-bit, October Update already. V 20H2 (build 19042.572)

Drivers: I am using the latest NVIDIA driver downloaded from Geforce Experience automatically (You can see the driver on GPU-Z screen above too). The driver version is 456.71 from Sep 30, 2020.
The Intel driver I didn't even bother to update at all: On GPU-Z, It shows that it is the 20.19.15.4416/ Win 10 64 from March 23rd, 2016. I just let Windows pick this driver automatically.

Thank you for the info on the temps! I asked because my gtx 880m ran in pretty hot temps and I think that this was the cause of its death in the long run. But this model (GTX 880M) is known of being trash. I just found out after buying the laptop . The gtx 980M seems to be a really great architecture improvement (Maxwell chips).

For the backplate, I remember that I had a problem with fitting the screws for the backplate, you are absolutetly right. What I did was: I went to a general electronics store and bought the correct screws to fit the GTX 980M with the backplate. At the Aliexpress purchase, I only recieved the gpu itself, a thermal pad and 4 screws. When I had this problem, I just bought 4 new ones that were capable of fitting the backplate.

Please, If you have any further questions, let me know and do not hesitate to ask. I know how it feels to look for answers for old rigs like ours that are still completely fine for everyday use and even some gaming, and not having a good amount of info.

EDIT: I had the same original BIOS and EC versions that you did (from PC Specialist, where I bought the P170SM-A) And have the Performance Scaling option turned on (the default IIRC).

 

@Shin Falcon Thanks for your reply !

You're right, there are probably strong similarities between those Clevo models, however I'm pretty sure you have a 15” screen, which it still a difference. I've had enough problems with BIOS upgrades and don't want to take more risks by flashing with a BIOS/EC from a different product.

Your 980M looks very much like mine ! 10DE-13D7 is probably the PCI vendor-product code. Right now I'm on Linux with my 980M deregistered from the kernel PCI database so I can't check but I remember for sure it starts with 10DE since I've already made a search with lspci filtering with the 10DE vendor code (10DE=Nvidia) and found it.

I don't think the problem is with the card ID however, since as soon as I had the GPU installed in the laptop (and before flashing the BIOS) I booted Linux and tried a 3D-intensive game (namely The Signal from Tölva), and it worked well ! It's only after flashing my BIOS that now it doesn't work on Linux. On Windows, I installed the Nvidia driver after the flashing, and it was recognised, I had no need to modify the INF.

Concerning my problem on Windows, it may just be a matter of reinstalling the Intel + Nvidia driver, in that order. I may try a BIOS/EC reflash too, with the one provided by Prema (from the BIOS thread mentioned before) instead of the ones directly from Clevo. While it's supposed to still be the stock BIOS, only repackaged, who knows, maybe it will do better. As a last resort I will flash my BIOS/EC to previous versions. I will try reinstalling the drivers on Windows today... I'm not much on Windows nowadays since I tend to favour Linux, except when it comes to a few games that are only on Windows 

Your KB backlight problem may be due to the fact that recent BIOS updates seem to have changed compatibility with more recent versions of the Clevo Hotkey program (as noted in the changelog). While I didn't update my Hotkey app, I noticed the KB backlight management seems to still work, but it doesn't deactivate the touchpad like it did in the past, I to have to type Fn+F1. On Linux, the backlight management still works like before.

My Keyboard has the Windows & Fn keys in the correct places, (Windows on the left, Fn on the right). That being said, your KB must be different than mine, since if I'm correct you have a 15” laptop, hence probably no numeric keypad...

 

@victor_alves Thanks very much for your detailed reply ! With all that info in hand, I should be able to have something working without having to downgrade my BIOS.

It's too bad you're not on Linux however, since for the moment I have options that may solve my problem on Windows, but it's not as clear on Linux...

But anyway, now is the time I boot Windows and try uninstalling/reinstalling the Intel & Nvidia drivers. I don't think I will upgrade Windows to 20H2 yet since I remember that time not very long ago when MS rolled out a bad upgrade that wiped out many a user's files. So now I tend to wait until MS forces the upgrade on me...

On your BIOS setup you say you have GPU Performance Scaling activated, for me it was off by default. I tried with it on but on Linux only, and it didn't change anything, maybe I will try on Windows too, but I don't think the problem comes from here. But who knows, that's still an option to keep in mind... If it was on by default for you, that may be the sign that the Prema-made BIOS has some undocumented differences with the Clevo-provided one.

 

Yes, i did have a 15.6" screen, and i'm pretty sure it's the only difference with hardware. Of course, chassis is not the same, and as you said, the keyboard. BTW thank for answering my questions!
For the vendor ID, 10DE-13D7 only means it's a 980m, the vendor ID is next : Mine have 1558 5106 with version 84.04.85.00.0D. I'm absolutely sure 1558 is for Clevo where MSI have 1462.
I can confirm Prema Bioses have some hidden option activated and some default settings changed.

EDIT : You said you update your BIOS using one from the tread you mentioned, but take care, they are all stock! Even if the ReadMe say "Prema Mod", the first line say "STOCK BIOS", same for the BIOS name. Mine come from this tread too, and i had a Prema Bios before this just to test if it can correct the keys for my KB. In my case, now i'm no longer able to change memory settings and i'm stuck at 1333Mhz with slow timing because stock BIOS don't have the Chipset menu activated...

 

The -a made some minor connector changes and officially supported an extra mSATA slot. In terms of things like power capability and heatsink mounting they are the same.

 

@Shin Falcon @Meaker@Sager Thanks for your replies. I have been occupied by other things, but of course I'm still looking for solutions.

I finally installed the very same BIOS @victor_alves did. I'm using the same Intel and Nvidia driver versions too, however it still crashes ! I managed to take a GPU-Z screenshot a few seconds before Windows crashed, here it is :



Things look very similar to victor_alves' config (including the same vBIOS version) except for two things :

• The bus interface : Victor has PCIe x16 3.0 @ x16 1.1
• The driver type : mine seems to be based on the newer DCH framework (Declarative Componentized Hardware)

I will have to check why I did get this, and possibly install a non-DCH driver. Concerning the bus interface reading, it seems to be related to the actual bus clock, in that sense Victor's reading is more consistent since his card probably wasn't doing anything at that moment. Mine wasn't either, thus should have displayed the same. I suppose it was busy crashing...

Anyway that doesn't answer the fact that it crashes equally badly on Linux, which makes me think the problem is unrelated to the driver type. After all, it would be really strange that an up to date driver would make the system crash this badly.

Ah another question however, do you have EFI Secure Boot activated ? I do use it. To check if SB is activated (i.e. enabled and system mode is "user"), you can either :

• Check your BIOS' security tab, and see if it says, well, that Secure Boot is enabled and system mode is "user". If it's "setup" that's not good, that probably means the platform key (PK) isn't defined, hence it doesn't perform any signature checking on the EFI executables, and the OS won't perform kernel-mode SB signature checking either. You should then set the PK by any means at your convenience, the simplest and quickest being to set Secure Boot mode to Custom instead of Standard, populating PK with the default machine-provided key, then setting back to Standard.
  
• On Windows, launch msinfo.exe, in the msinfo window there's some info somewhere about secure boot status. However, I'm not 100% sure about this, but if you have a Linux-Windows dual boot it's possible that the MOK system interferes and makes Windows think wrongly that SB is activated. At the very least the MOK system has been reported to prevent Bitlocker from working correctly. That's why I don't use a traditional Windows-Linux multiboot, but instead choose my OS from the BIOS boot menu.
  
• On Linux, some distributions use the mokutil program, but the MOK system's concept of Secure Boot is quite different, since it's a software shim that's applied at runtime over the machine's Secure Boot system. I suppose however, that the mokutil --pk command displaying the plaform key could give a hint : if it doesn't return anything, then there's no Secure Boot. mokutil --sb-state is not a clear way to know if secure boot is activated, as it can be controlled independently of the actual machine's Secure Boot.

So if you don't have Secure Boot, maybe that's the difference. So few people seem to use it in practice, that maybe there's a bug that slipped through unnoticed.

Your advices are welcome ! But as time goes, I'm considering more and more seriously a BIOS downgrade. After all, the only time when I managed to have my GPU working was with my computer's original BIOS. But why the heck it works for both of you and not for me, is beyond me right now...

 

Hello,

For the bus speed, yes, if it's locked to 3.0, may be 1.1 at idle. And i'm with you about the driver, pretty sure it's not related to your driver type.
I'll check the EFI later this morning and i'll edit this post to say. @victor_alves setting can be a clue too and have the same setup as you.
Sometime, there are problems we just resolve without understanding what's wrong : some weeks ago, my laptop didn't display anything for 30-40s before bios post, just the power LED ON, not event the screen was ON. And then boot like normal and performing the same as usual. Nothing was changed on the P150-SM. I did try to load default bios options, i updated my bios from 1.03.04 V2 PM do 1.03.05 stock, without change. What worked? Removing any supply source including the CMOS BATTERY, waiting 5 minutes and all was normal again! Just a full clear CMOS! Why? I'll never know!

BTW did you test with another GPU?
The bios downgrade should be easy if you take the file from the unoficial Clevo Repo, it's like a one step flash. Remember, UEFI and intel anti thief must be off before flashing.

 

THE CMOS BATTERY ! I will have to check this indeed, I semi-consciously had that go through my mind but never really thought of it, because in my previous Clevo laptop (M571RU-U), I wasn't able to get to it easily. But where did you find it ? I suppose it wasn't under the GPU since you would have to repaste it otherwise, which would be a pain. I will check the service manual... My laptop being bigger than yours, it should be easier to get to. Moreover, maybe there's a CMOS reset pin somewhere, that would make things more simple.

I downloaded the standard non-DCH driver from Nvidia, I will test that too.

I didn't test another GPU as I've no intention of buying more than one. Of course there's the possibility that the one I bought is defective, but then why did it work before the BIOS upgrade ? That being said, there may be a difference between mine and the one @victor_alves which would be in hardware, since we have the same vBIOS...

The problem with the BIOSes from the unofficial repo is there are probably some missing, as they only provide a handful of EC/BIOS combinations, and most notably it's missing my original versions. That being said I'm considering a progressive downgrade, the idea being to have the most recent EC/BIOS combination that works.

About the problem you had with the 30-40s wait before POST, I had it too, but that was at the time my dead GPU was inside, and was clearly because it was unresponsive to the BIOS initialisation. I didn't have the problem any more after I replaced it with my 980M.

 

Hi,
Check under the keyboard, you will find a small connector under the keyboard flex cable. On my P150SM i have to remove the front speaker cover / power switch bar, en then unscrew 5 screws. No need to remove completely the keyboard, just flip it but be carrefull not to break one of the 2 flex cables, i don't know how yours is built exactly but can't be very different since we have same board. The service manual can help you a lot on steps to follow to access it.
Remember that our laptops works in a "strange" way and you may check this tread at post #7 by @Meaker@Sager : http://forum.notebookreview.com/threads/quadro-p5000-with-clevo-p150sm-a-lvds.814127/
So the IGP can work "alone" for low display task and even if your dGPU is correcly detected by the bios, it can boot and start. The iGPU works as a "display switch connector" and if you run a game for exemple, the dGPU will send what it want to display to the iGPU. You can understand that if your dGPU have problem but can be detected by the BIOS, your laptop will start.
If you still have your 880m, you may give it a try and see if you can reproduce the same crash with it.
Another option you can try is cleaning the dGPU connector with some isopropyl ethanol (after removing ALL suply of course and both the card connector and main board) and see if it's working better.

For now and after all you did, i begin to think your 980m is faulty.
I searched differences between 880m and 980m and foud that there are 2 "grey block" on the top left of the card. you may check if your cooling pads are fine make proper contact. Maybe it's just a bad VRM cooling?
BTW, i still have a working 770m. Too bad we have to stay at home for some time now but if you want it for testing purpose, send me a PM

 

Inductors I guess you mean? The tall ones? You do need to watch those in particular with heatsinks as they do get in the way.

 

@Shin Falcon @Meaker@Sager Thanks again for your replies.

Today I helped someone moving, despite the nation-wide containment measures that just resumed in France, so I didn't do much, however yesterday (thursday) I found where the CMOS battery was by looking at the MB picture in the P170SM-A service manual. So I was able to perform a CMOS reset, and the system clock stopped running for the first time in more than 6 years. To be sure, left the battery unplugged for 15 good minutes.

I had high hopes, but unfortunately it didn't change anything, still almost instant crash.

I'm still not convinced my GPU is really malfunctioning, since I tried the very day I installed the card and it worked well for 30 minutes of gaming at least. During that time I had my GPU's temp under scrutiny and everything was fine (don't remember the temps precisely, but I'm sure it never reached 80°C because that always triggers a special notification in my brain). But since I noticed my GPU wasn't identified by the BIOS any more, I thought maybe it could be good that I upgrade to the latest.

I don't think it's an overheating problem for two reasons :

• If it's the die I would have seen it without a doubt, since I was monitoring the temperature, and I suppose the sensor lies there.
• If it's not the die, it would probably not heat as fast, hence take some time before crashing, which isn't the case here : it crashes within seconds.

I could add a third reason, the way it crashes. On Windows, MS has a system to detect (and recover from) a crashing GPU, that's generically called Timeout Detection and Recovery (TDR). I've experienced my fair share of TDRs on my 880, and it does the job, it recovers. Being on an optimus system, the associated program usually crashes along the way, and the GPU is powered down upon driver recovery. The day my 880 definitely died there was a freeze indeed but I could see the OS wasn't completely halted (only the display was affected), as there was still some disk activity. After that (on subsequent boots) my GPU just failed to initialise and the driver didn't load, which would not prevent the OS from running. In my current case, the whole OS seems to halt completely : no more reply to pings, no more disk activity, all of a sudden, within seconds.

Silly me... If I could go back in the past I would not touch that damned BIOS, that's for sure now. I'm surprised both of you could upgrade their BIOSes without a single problem. I must have been cursed by the BIOS gods, since upgrading a BIOS never did any good to me. That may be the last time I do such a thing in my whole life ! Now I will have to downgrade as carefully as possible, hoping not to brick my system, as that operation is generally unsupported. It's even explicitly discouraged in the service manual :



@Shin Falcon I appreciate your offer to loan me your 770, and I will not rule that option out, but it's not like my 980 never worked, there clearly has been a before and an after BIOS upgrade. I think my best bet now might be to downgrade. Maybe I will contact my reseller before, in case they could bring me the original BIOS 1.03.03 and KBC/EC 1.03.02 I had originally, since it doesn't seem to be available in the official Clevo repositories, or the repo mentioned in the old Clevo BIOS thread on this forum.

Oh and btw, you didn't tell me if you have Secure Boot activated. Do you ? Or maybe you don't even use UEFI and use the old MBR system instead ?

 

@Meaker@Sager Thank you for clarifying, yes, it's inductors. It's not always easy to find an english word when it's not your native language
@NovHak Sorry, i think i had to stay at home like you but my boss said we can stay open for some reasons. We will not debate if it's usefull or not
My UEFI is ENABLED in the BIOS.

I can understand your fear of flashing a BIOS but really, as soon as you have the good files, it can't turn wrong
I flashed a lot of BIOSES and on my P150SM even "played" with it using AMIBCP to unhide some options and flash it under windows with AFUWIN64. I do not recommend you to do this !
In your case :
- Create a USB flash disk with RUFUS, it have an option to make a FREEDOS boot disk.
- Download the archive you want from the repo, really it's a lot easyer than asking you reseller.
- Extract files in the root of your USB disk not inside anyfolder
- Go to the BIOS and disable UEFI the, save and reboot
- Tap F7 at the BIOS boot screen to select your boot device and select your USB key
- Freedos will ask you to select your keyboard layout
- type update.bat then enter
- Confirm to flash the BIOS
- Wait for the program to complete erase/writing/verify of all blocks it can take a minute
- Then another screen ask you to flash EC, just press y and confirm.
- Once done, you'll see your laptop will power off in 3s, just wait...
- Once off disconnect the power supply and wait 20s
- Connect the power cord then power on the laptop and wait. It may power off then on, just wait for it to display something.
- Enter BIOS and load default, then change what you want. Don't forget to enable UEFI.

EDIT : UEFI is ENABLED but SECURE BOOT CONTROL is DISABLED

 

@Shin Falcon Thanks for the info ! So you don't have Secure Boot enabled... I will have to check with Secure Boot disabled too, but I would find having to disable it definitely an annoyance...

However, I found a difference between the flash files you used and the ones I used ! I thought the stock BIOSes packaged by Prema were exactly the same thing as the Clevo provided ones, except they are packaged to flash BIOS+EC in one go. I was wrong... The Clevo-provided BIOS additionally flashes something else !

Let me go into more detail : additionally to P17SMA.11 that's exactly the same byte-to-byte as P17SMA.311 found in Prema's package, and for the EC, eP17SMA.09 that's the exact same as prema's eP17SMA.39, there's another file named P17SMA6M.11, that's bigger and flashed before the BIOS, with another tool named fpt.exe.

After searching a bit, I found it's an Intel PCH (Platform Controller Hub) ROM, in other words the "chipset" ROM, except it's only one chip. Prema's package doesn't flash this.

I will see if updating my chipset driver will help. I could try getting an older Clevo BIOS and flash only the PCH part too. But now at least, we have a clear difference between what
@victor_alves and you did, and what I did. Heck, so much for getting the official Clevo BIOS package, this time I should have chosen the shortest path !

As I said before, I don't believe it's a GPU problem since it was working before the BIOS update... or should I rather say, the PCH ROM update.

 

Unless it's got an incompatibility with that specific vbios?

 

@victor_alves uses the very same vBIOS with his 980M, and updated his BIOS & EC to the same version I did. The only difference between us is that my chipset ROM has been flashed too.

There's something very wrong with the crashes I experience, I noticed when I power back on after a crash, sometimes I have caps lock activated, sometimes Fn+F1 to disable the touchpad doesn't work at all. A problem with the chipset ROM could explain those things.

EDIT :
@victor_alves @Shin Falcon Could you tell me what your Intel Management Engine firmware version is ? You can see that on your BIOS setup main tab : I think you'll get something different than me ; I have 9.1.0.1120, but I guess you will have something starting with 9.0. I've read somewhere that downgrading the BIOS while having ME FW version at 9.1 could pose problems, because not all BIOSes can work well with 9.1 versions...

 

Hey! Nice find! I did see these files some times ago but didn't really wonder what's inside. Looking at an original Clevo package did have these files, i remember trying to flash it without success.
From what i've found, this file may be related to the ME firmware, not sur if it's flashing something else. You can give it a try but not sure it'll do something.
BTW my ME version is 9.0.20.1447.

EDIT : Tried to flash the latest bios from Clevo WITH the 6M file and doing exactly what said in the readme.txt (and not redme1st or readmeec). So fpt.exe flashed something but there was an error at first saying the file is too big then it flashed some small files to some adresses.
At the end, it says something like "files are the same" and my ME version is 9.0.20.1447 so nothing changed.
At last i have to flash back my modified main BIOS to unhide option.
Since i already did flash these files before, i think it's why nothing changed.
BTW, are you sure you don't want to try my vBIOS?

EDIT 2 : Flashed a full Clevo P150SM-A (so with the 6M file). And now the ME firmware is the same as you, 9.1. But there's something more! My keyboard backlit is now "working" even if i can't change color, i will make a try later with the latest CCC since i still have the 8.0.150.
980m installed with an old driver from 2017, updated succesfully with the latest one (modded).
Everything is working fine for 1h now.

EDIT 3 : My bad... EC version is still 1.03.01... That's why my KB is working...

 

@Shin Falcon I managed to flash a 6M file from a 1.03.02 Clevo BIOS package and now I'm back to ME FW 9.0.20.1447, however I had a warning about the flash file being bigger than the flashed region, so maybe something isn't really back to the previous situation.

I then tried a few things, but I didn't see a big improvement. The GPU still crashes, maybe it takes a little longer to do so, but maybe that's because I tried the non-DCH driver.

From now on, I'm not sure I will continue testing on Windows, as it takes more time especially when it crashes before I can't do anything, now that F8 works once per century I have to wait for the OS to crash a few times before it finally decides to go into maintenance mode, and choose safe boot. Now if that's not bad design or simply carelessness... I know I could configure the bootloader so that it pauses before booting, but why the heck don't they do anything about that F8 ? Anyway... on Linux I can quickly see if things are going to crash, and most of the time reboot cleanly before it's too late.

Next things I'm going to do (until it works again, hopefully) :

1. Flash BIOS to 1.03.08 & EC to 1.03.08
2. Reflash BIOS 1.03.08 along with ME 9.0
3. BIOS 1.03.05 & EC 1.03.07

Then I'm not sure, because I don't want to go back to EC 1.03.01 since it's older than what I had originally. In the meantime, maybe I'll get the original versions I asked from my reseller (LDLC).

My goal at this point would be to get back to the configuration I had when everything was running OK, then if the GPU doesn't work I could consider it's faulty. At least I would like that it powers on/off the same way it did before on Linux, when I simply had to call the ACPI _ON method + scan from the PCI hub to use it, and _OFF to power it down (either directly or via bbswitch, after having removed the nvidia modules of course) + remove the PCI device from the kernel.

Currently the _ON and _OFF methods don't work well, using them seem to make the GPU unresponsive to PCI scans, even when it's on.

I have to rescan twice from the PCI hub where the GPU is plugged in, the first time powers it on, the second one makes it detected by the kernel (and the modules are loaded). One advantage with the current BIOS is that the GPU is automatically powered off when Linux boots if the Nvidia Prime Intel profile is selected. With my original BIOS I had to call the _OFF method... but it wasn't that much of a hassle, especially compared to now where the GPU simply doesn't work.

 

I had this warning too, but in my case it did update the ME firmware so maybe it's OK.
About your Linux command, i don't know why you do it that way, i guess optimus is not working and so you have to do it manually?
I think you can directly try 1.03.05 & EC 1.03.07. But, really, i don't think it'll change anything.
Did you try to remove the card, clean connectors, check thermal pads and fit it back? Really, try it, can't be bad.
Using Linux, i remember some options about ACPI on the Prema bios, but really, i don't understand what it mean, and it's seems to be disabled by default.

 

@Shin Falcon On Linux I have three choices (using Nvidia Prime) :

• All graphics on the iGPU (prime-select intel)
  
• All graphics on the dGPU (prime-select nvidia)
  
• Graphics on the iGPU by default, on the dGPU on demand (prime-select on-demand)

The three solutions work, however, at least for my 880M, the drivers don't support switching it off when it's not used (iirc, it works only for GPUs on Turing architecture and newer). The dGPU remaining powered even when not used, I consider the on demand solution not that interesting. So :

• When I'm not using the dGPU, I activate the intel profile, logout (which normally automatically removes the nvidia modules) and back in, call the ACPI _OFF control method on the dGPU so that it shuts down (or use bbswitch to the same effect), and remove it from the kernel PCI devices database, to prevent subsequent hardware probes from powering it back on.
  
• When I want to use the dGPU, I activate the nvidia profile, call the ACPI _ON control method that powers it on, launch a PCI scan on the PCI hub where the dGPU is connected (which automatically loads the Nvidia modules upon detection), log out and back in.

By doing so, I can switch it on and off as often as I want, without having to reboot.

Concerning my current problem, switching to BIOS 1.03.08 & EC 1.03.08 had the effect of bringing back working ACPI _ON and _OFF methods, which I prefer. However, it still crashes immediately upon module load.

I will continue as I said, so now I will reflash B08 and ME 9.0, but you may very well be right, I'm more and more afraid that it's not a BIOS problem after all, but a GPU problem. What still gives me some hope is that once I had it installed and before flashing I was able to play for 30 good minutes without any problem, then shut down and unplugged my laptop, and went to bed. Why would it fail completely the next day, working only a few seconds at most ? Considering I flashed my BIOS just before retrying, it could arguably be a BIOS problem.

Additionally, I took note of the Nvidia driver Xid error when it crashes, and it always gives my Xid error 62. According to this, it means Internal micro-controller halt, with possible causes (according to Nvidia) being : HW Error, Driver Error, and Thermal Issue. That being said, a driver error encompasses multiple causes, among which BIOS related problems, as drivers often make use of ACPI, for example. If it's a thermal issue, it's not on the Nvidia chip since during my 30 minutes play session I carefully monitored its temperature and it was really OK.

If nothing else works, I will indeed open my computer again and recheck everything, maybe remove the shielding plate that was attached to the 880M but which according to @Meaker@Sager can be put aside. Did you have to attach a shielding plate to your 980M ?

I don't think using your vBIOS version would help me, since iirc it's not a Clevo BIOS, which is probably why you were forced to use a modified INF. Moreover, @victor_alves has exactly the same vBIOS version as me, and last but not least, at the beginning my GPU was working well... but you may still be right, maybe the reseller flashed the card with a Clevo BIOS despite it being, say, an MSI card, in which case the vBIOS could be inadequate, and maybe put an excessive strain on some peripheral components.

 

I can understand your frustration and as you, for now, i'm not understanding why it worked for 30 minutes then nothing.
There is one thing you can't monitor on a 980m, it's the VRM temp and how well it can be hot. On mine, at first, it was a problem since the card is not designed to be limited to 100W and the VRM cooler was way hotter than the one on the GPU and having heavy power throttle. What i did is simple, make a backup of the BIOS and keep it, and modified a copy of it to lower voltage and see how far i can go with max boost on the lowest frequency. Doing this way, no more power throttle, lower GPU temp, and using my finger i can say VRM is colder too. Thermal pads are VERY important and there is differences between a 880m and 980m as said before with at least one more inductor that need to be cooled. There are different bads too, some 0.5mm and some 2mm. Really, give it a try, in one night, a too thin thermal pad can reduce and will not make proper contact anymore.

About the shielding plate, no i don't have it and used the "X" plate from my 770m so the back of the card is not cooled. I was told this shield is not needed and recommanded on our board and it's just to cool the vRAM chip under the card. Maybe try without it?

A BIOS is like a driver in some way : there are updates to correct some bugs or add some optimisation. And even if you think my vBIOS is not a CLEVO, just check the .inf in NVIDIA drivers : there are at least 6 or 7 980m listed in the CLEVO inf so at least 6 or 7 980m begining with 13D7 10DE (980m) and SELLER ID 1558 (CLEVO / KAPOK). Only what next change. Why? Maybe because there are clevo resellers around the wolrd that have their own last digit (5291 on mine for exemple) or maybe for some revision of the board design.
About this, check you're BIOS build date and you'll see that mine is newer. What changed? I don't know... But remember not everybios for a said card can work on every card and i already had a card that was working fine on desktop and crash on a heavy apps just because the BIOS did not have proper timing for vram chip at high frequency.

So all this to say, the error you have in Linux ask you to check exactly what i said you : check your temp at first and try something you have not done yet : change your vBIOS. If after this we still don't have it to work well, then try to downgrade your main BIOS.

 

Indeed I wasn't very confident with the thermal pads placement. For the inductors I had enough thermal pads, but for the VRAM thinner ones were necessary, which I didn't have, so I reused the ones from my 880M (which luckily had the same VRAM placement). Hence I think there's no problem with the VRM (I remember I covered a few mosfets too).

I wasn't familiar at all with MXM replacement so I had a look at a few videos, and heck, thermal paste blah blah blah, but none mentioned anything about the thermal pads placement. So, the inductors, nearby mosfets and VRAM chips must be covered, anything else ? In the end what exactly needs cooling ? Don't tell me the whole card needs, since I've not seen any configuration with thermal pads everywhere.

Anyway, I'm going to buy a stock of thermal pads of differing thicknesses (0.5/1/1.5mm). Being on the countryside and with the current lockdown I will have to order this online, which will take some time, and even more so precisely because of the lockdown. That's very annoying. In the meantime I suppose the best I could do would be not to use my GPU !

Concerning a possible vBIOS update I'm not sure it's a good idea, precisely because of those freezes seconds after using the GPU, wouldn't I risk additionally bricking it because the flashing would hang half way ? Note that simply monitoring the GPU temps make it crash after a few seconds, I don't even need to perform any sort of graphics task.

 

No even with over volting of core and vram for back plate.

 

Then its days are probably counted !

The things that I possibly did wrong when installing my GTX 980M are :

• Reusing my 880M's thin termal pads to cover the VRAM chips.
• Not fastening the shielding plate to the card because the 880M's screws didn't fit any more. However it seemed to me that it was fastened enough, being under the card that was already fastened to the chassis with those two hex screws, plus the three screws fastening the lower heatsink which iirc is fastened to the chassis too.

 

Thank @Meaker@Sager for confirming for the back plate. I don't have one and temps are good this way.
@NovHak so you may disassemble your 980m and remove this backplate, then check the thermal pads according to this tread (even if it's not a 980m), see post #13.
Take care of the 2mm on the top of the card (covering the core VRM) and 1mm on the right of the card.
You'll find 2 inductors instead on one on the top left that need to be covered.

Hope nothing breack because of heat!

 

Contact is the most important think to VRMs first and foremost, then the core and then capacitors.

 

@Shin Falcon Yes I saw similar pictures already. Like the poster said, I suppose I should ignore thickness measurements and check myself. I only have one thermal pad sheet of one thickness, so I bought three 50×50mm² sheets with thicknesses of 0.5, 1.0 and 1.5mm. I guess that will be enough.

I could try doing with the small sheet I already have, but if I don't want to reuse old pads I would have to peel the new ones to get the correct thickness, which I'm not sure is the best of the ideas. I hope the postal service won't take too long !

@Meaker@Sager When you say VRM, I suppose you mean the inductors ? Because correct me if I'm wrong, but the capacitors are also part of the VRM... and am I guessing correct that you're speaking of the Nvidia chip when you speak of the core ? If would be difficult to forget this one, but I'm pretty sure I did it correctly at least, since the sensor must be there and I carefully monitored the temps already. But in that case, it seems to me you forgot the VRAM, or is that what you mean by core ?

 

No I mean, VRM as in voltage regulator module. The set of black chips that gets hotter than the GPU core.

Maybe you are thinking of a power phase as a whole?

 

Okay, I think I get it. I didn't yet reopen my machine as I'm waiting for new thermal pads that should come by postal mail soon. I don't think I forgot covering some parts, since I looked at what had been done on my 880M and took inspiration from that... but I certainly didn't realise how much heat was dissipated by those non-core components, and probably didn't check well enough if effective contact was made. Anyway, now I don't see any other reason that would explain my "current" problem !

 

To explain a bit more what @Meaker@Sager said, you have 2 essencial parts to cool. I'll explain according to this 980m exemple.
- The Inductors, the "BIG" gray square with 2 large pins. You can find 2 of them horizontally placed on the top left of the card and 1 same with a smaller on vertically placed near the top right of the card.
- The VRM (voltage regulator) can be VERY hot (cut down temp of 125°C and average working temp of 100-110) and 3 of them are located on the top of the cards (its like a rail) for the core, and 2for the memory just on the right of the smaller inductors. You may have a 1.5 or 2mm pad covering the whole top of the card and a small one for the 2 on the right.

Excluding the power supply of the card, you have to cover the 8 memory chips with 0.5mm

I really hope your problem is there!

EDIT : your 880m only have one inductor on the top left so at least you may have one not properly covered. Remember for the VRM the pad must cover the entire chip.
Even waiting for your new thermal pads, you can open and see what it look like and compare. You can try posting a picture of your card too to see if something is brocken.
"current" problem, what a good joke

 

Inductors are less important than the capacitors, an inductor is just a coiled bit of wire.

 

Yes, you're right, but you made me wonder if i said something wrong.
The "grey square" i talked about are inductors or capacitors? I must know the correct word in english
BTW do you know what are the small black rectangles between both row of "inductors"? I wonder if i need to put thermal pads or not, i never seen someone saying to cool these parts, but i want to be sure

 

I can answer this, as I've seen much info about that, the grey squares are the inductors. I suppose if one was to open it (don't count me on that !), one would see a coiled wire inside.

When you speak of the black rectangles, you mean the ones between the two R22 inductors ? IIRC I had those covered by pads too, since it was done like that on my 880M.

Concerning the parts to cool I've passed much time scrutinizing this now. What worries me if that if it's a problem of one of the VRM mosfets reaching the cut down temperature, then how the heck could it reach 125°C only a few seconds after being started, with no intensive task being processed ? I'm afraid something has been broken, but that's possibly because I've been given a bad card. After all, like you said there are cut down systems, and except for those 30 minutes of gaming that went quite well, I've done nothing demanding on that GPU, I mean what, once at most I displayed the little turning Nvidia logo from the control panel on Windows, let's be serious...

I will indeed send a picture of the card once it has been removed, but it's probably not the kind of problem that will be visible. For now I'll wait for the pads to come before opening, and keep my GPU unpowered (hybrid graphics light off). One thing I appreciate with the BIOS & EC 1.03.08 is that on Linux, once it's been powered down it totally remains so. With BIOS 11 & EC 9, it was still blinking very shortly at times such as when logging on/off.

And don't get me started on Windows, where it will switch on from time to time with no apparent reason, and clearly it must be using it since that's when I get BSODs and/or complete freezes. The only way to avoid a crash is to disable the device in the device manager, but then it remains powered on, what an annoyance. Calling _OFF is more complicated since the Windows framework doesn't permit a driver to call arbitrary ACPI methods, it can only call methods that are child to where it's attached in the ACPI device tree. Attaching to root probably can't be done since the Acpi.sys driver is already there, so contrary to Linux, there's no generic way to do this, one would have to write a driver that attaches to \_SB_.PCI0.PEG0.PEGP, and call methods from that point. And I'm not even started on driver signing, which I'm not sure is even possible while retaining Secure Boot without paying $100 (which would be unacceptable), but I'm not sure about this. Note that I'm speaking of a driver for private use, not general distribution.

I have another question that may be better asked on the owner's lounge thread : Is there a way to keep using the P170SM-A with no dGPU inside ? Because I've heard that then it beeps and shuts down shortly. So either via means of a dummy MXM device, or a BIOS and/or EC mod, I suppose it would be possible ? That could be interesting so that if I have to wait before putting it back on, I would not have to reinstall my now defunct 880M...

 

Reaching 125°C for shut down condition or... Maybe a blown VRM? That's why it can be usefull to see the card, if you see a trace of burn near or on a VRM chip, it's a bad sign.
About your driver, it may be possible without secure boot by activating test signing. Maybe you can bypass this entirely by generating your own certificate and signing the driver with it, but i'm not aware of what can be done on this way.
About using your laptop without dGPU, it seems to be possible to desactivate it with Prema BIOS. Not sure if it'll boot without it, but you can try disabling it via BIOS.

 

You would need to ask Prema if he has an iGPU only bios he could provide (by default the system will always shut down in an ECU panic without a GPU).

 

@Shin Falcon and @Meaker@Sager Thanks for your continued support !

I've received the thermal pads and removed my 980M, for now I'm putting the 880M as a replacement. Now I can send some pictures (the links send to a bigger picture) :

1. The card as I found it just after removing it , with its thermal pads. I know what you're going to say so let's get that out the of way, one of the pads fell when removing the sink and just before taking the picture I mistakenly replaced it between the vRAM chips and the R56 inductor, but of course it was on the visible R22 inductor ! I wouldn't have cut an additional thermal pad the size of an inductor to just put it there. I reused the 1 mm pads from the 880M (vRAM and R56 inductor), and replaced the R22 inductors' and VRMs', that are 2 mm thick (hence 0.5 mm thicker than the old ones, which may have been a mistake).
   
2. The front of the card
3. The back
4. The card, repadded . I have a problem with this however, since I padded with the following thicknesses, but it doesn't work (the sink doesn't seem to touch the VRM and bottom vRAM row well if not at all) :
   • vRAM : 0.5 mm
   • R22 inductors : 1 mm
   • R56 inductor : 0.5 mm
   • VRMs : 1 mm

The factory configuration (with my 880M) seemed to be :

• VRAM : 1 mm
• R22 : 1.5 mm
• R56 : 1 mm
• VRM : 1.5 mm

So at first I thought, if I remove 0.5 mm thickness everywhere it should be better, since thinner pads mean better cooling (with contact, that is), like I said it doesn't seem to turn out very well. Now I'm hesitating between :

1. Use 1 mm for the vRAM and 1.5 mm for the VRM (or maybe even the remaining 2 mm I have), but then I won't have enough 1 mm in stock.
   
2. Add 0.5 mm everywhere (maybe even 2 mm for the VRM). That's more consistent with the 880M thickness setup and I may have enough provided I can reuse the two R22 pads for the vRAM.

It is difficult to make sure proper contact is made everywhere, and that's why I will probably choose the second option since it's the nearest I can find to an official specification on the matter, but your advices are welcome ! In particular, I remember @Meaker@Sager speaking of the capacitors, but for me, a capacitor has a cylindric shape and I don't clearly see one here...

 

For VRM use 2mm, then you can use 1mm for memory chips and 0.5 for inductors.
Watching at your 1st photo, the middle top VRM is not properly covered and one inductor is not covered.
Remember the cooler is not flat too!
About capacitor, there are 2 mains type : ceramics and chimics. The round cylender you talk about is the chimic type and it tend to break over time and can spread acid and/or explode. Usually used for high voltage. Ceramic is mainly used in CMS package, or like a small sound piece like a smarties (Or Skittles), used for low voltage and very robust.
For me it seems like your card is well covered, you can try to use the olds you have on inductors but properly cool VRM is very important.
BTW i did a mod to my GPU cooler using some Araldite (it's a type of strong glue, like bi-component epoxy) mixed with a piece of an old cooler i cutted with a dremel. The purpose was to use a 1.5mm pad between the GPU cooler and the large black cover of the laptop to make a better pressure on the cooler and it's working fine : my 980m now run at 1177Mhz @ 9.75V stable without going at more than 80°C after an afternoon of gaming and didn't see any power throttle

 

@Shin Falcon As I said, I know the top inductor isn't covered in the picture but it's just that I didn't replace it properly after it fell when I removed the sink. Concerning the middle top VRM, it looks like part of the chip is visible indeed, but I think it's an illusion and the metallic part that appears is not part of it. Anyway, the 4th picture shows what I intend to do in terms of card coverage.

Of course I know the cooler isn't flat, otherwise I would need much more thickness on the vRAM !

Since you said I can use the old pads from my 880M on the inductors then you're validating their thickness, so I think I will choose option 2 (but still will use new pads) : 1 mm on the vRAM & R56, 1.5 mm on R22, and 1.5 or even more likely 2 mm on the VRM. Note the R56 inductor is slightly higher than the R22s, with the sink's surface not being different, which is why I put a slimmer pad there (the same that was done in factory btw).

I will likely reinstall my 980M tomorrow, but I'm not making myself many illusions, it probably won't work and the card is dead... but we'll see !

Regarding your GPU mod, I don't understand clearly what you did, but anyway if I had a working 980M I would be happy already... and considering I won't do any form of overclocking, and I always carefully monitor the temperature, it probably will be enough for me !

 

With my 980M because I ran it at a much higher clock I got the additional 3 VRMs soldered but that can be tricky if you don't have a good reworking shop in your region.

 

@Meaker@Sager You mean you have 8 VRMs ? I suppose your cooling system must be top notch