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

@Meaker@Sager I'm skilled with soldering but not with CMS ones. I don't want my card to overheat so that why i just made a "light" overclock editing the boost table and boost limit so it run @1177Mhz stable with table n°41 if i remember well. My ASIC quality is 69.8% so voltage is around 0.975V and nearly never power throttle for just one second and down to 1152mhz. Same for memory, i see lot of poeple running at 1500Mhz but mine is not even stable at 1450 so i let it @1375Mhz. I want to keep my laptop as cool as i can so i'll not go further Maybe in one year or two i'll change it for a 1070 and mod it to keep in 100W.
@NovHak since it's hard to explain in english, i'll try to say it in french for my mod. What follow is a total HS! It's a french translation of what i said in post #47

FRENCH HS : Comme je l'ai dit, j'ai fait un mélange d'Araldite (une sorte de colle un peu comme l'epoxy à 2 composentes) mélangée avec de la pate thermique pour avoir quelque chose de thermiquement condictif. J'ai collé au dessus des 2 heatpipes un carré métallique d'un dissipateur provenant d'un vieux PC portable pour augmenter un peu la surface de dissipation et surtout me rapprocher du bord du cache en plastic qui recouvre les composants, ce fameux cache noir en plastic. Ensuite j'ai ajouté un pad thermique de 1.5mm d'épaisseur au dessus de ce que j'ai collé histoire d'avoir une forte pression sur le dissipateur grâce à ce cache une fois fermé. J'ai perdu 7°C et ma carte monte du coup rarement à plus de 80 (81-82°C après 2h de FF15 MAXIMUM). Par contre, c'est très moche! Au passage, j'ai aussi remplacé le woofer mais là je ferais un post en section dédiée avec photo dès que j'aurais un peu de temps.

BACK in english : i can understand what you think, but your card seems to be OK, there doesn't seems to by any damage but i can be wrong. At least i really hope you so. You may re-use some older pads and keep the same configuration as the default, just add the one missing, it's seems to be fine looking at your pics. On mine, i didn't change anything from my 770m so the pad for the 3 VRMs on top is a long one covering all the way from left to right even when there is no VRM soldered to the PCB. I just added one 1.5mm on the inductor from the same old laptop i used for my cooler mod

If after this your card is still not working, the only thing left is flashing your vBIOS. I tested the same vBIOS as yours on my 980m and it did work but only on safe mode. I didn't make much effort, pretty sure it's just a driver problem but once windows loaded, i just have a black screen until i force the laptop to shut down. I flashed back my bios after this.

Remember to ask Prema if a iGPU only BIOS exist

And don't worry, i think it's normal to try helping each other and if i translate correctly : "If you know, share, if you don't know, ask!"

EDIT : About adding VRM, no it can make the card works cooler! If you want 100W and ask this to 3 VRMS, they'll work harder (100/3) than if they were 8 (100/8). This way, i understand he can ask more power/voltage from them for a better overclock. But, as you said, a stock 980m is already still fine today for gaming @1080p.

 

3 VRMs is more efficient at stock, 6 just helps when really pushing it, also 6 can impact your memory clock as it's actually a bit more noisy.

 

It was running them in my P570WM, I soldered the extra VRMs to help stability when over volting both core and vram.

 

@Meaker@Sager did you soldered it with a standard soldering iron or something else? You make a good point talking about noise, and it's logic since there is more power chips soldered to the ground. Maybe filtering the ground can help this but maybe it need to redisign the card... Bah and it'll not help our friend here

 

I had a professional BGA reworker attach them, you need the contact on them to be spot on.

I never got around to looking but upping the bypass could certainly help.

Bear in mind this was over volting the memory and getting it the 1850-1900Mhz region and then being restricted to 1750-1800Mhz after the power mod.

 

@Meaker@Sager It's still a nice overclock! And just a 100Mhz less on memory may not have a huge impact on these card (already 256bits bus). I remember back in time when i edited vBIOS for my Radeon X1900 to change timing settings and find what works best! I feel like it can't be done anymore or it's well hidden within the BIOS code. Sometime you can gain much more for lowering timing than upping the frequency.
@NovHak Don't be affraid if we're talking about overclocking here, i did pray a lot for your card to work. It will be back working!

 

My 780M cards went all the way up to 1950Mhz memory clocks which was pretty crazy.

 

Oh yes, this is crazy! I don't remember playing much with my 770m since i replaced it with my 980m 1 week after i received the laptop.

@NovHak no news from you?

 

@Shin Falcon @Meaker@Sager Sorry for having been silent this long ! There were two reasons for this, one being I had to help someone that's teaching music online with her setup, since some students only accept remote teaching nowadays... and the second reason is that I had lost some motivation, because I'm afraid the card is dead, but we'll see !

I've just put my 980M back into my laptop but didn't try it yet, I'll do it any time now.

Finally I've put 2 mm on the R22 inductors and 1.5 mm on the R56, since after having a second look, the original pads seemed thicker than what I intended to do in the first place ; moreover, I had doubts that effective contact was made between the inductors and the sink. I noticed something positive, after having applied the sink once and removed it, two pads that were on VRMs remained sticked on the sink, which obviously means there was contact. Btw, on the VRMs, instead of using the 2 mm pad I have which I think is of lowest quality (and conductivity), I stacked 1.5 mm + 0.5 mm of my 6 W/m.K Arctic pad, being careful that everything remains clean in the process. Now I'm hoping for the best !

In case it doesn't work, well, why not flashing the vBIOS indeed, but if it was a vBIOS problem, then why would dit work OK during 30 minutes after the first installation ? Anyway, I was fearing that the microcontroller crashing in the middle of the vBIOS flashing process would make things worse, but what could be worse after all ? That recalls me of a long time ago, when I flashed my old 8800M GTX's vBIOS (on my M571RU-U) but had to reflash the old vBIOS back since the temperatures had become significantly higher.

By the way, I suppose there's still only a Windows version of nvflash, no EFI or Linux version ?

@Shin Falcon After your explanation in french, I now understand what you did ! Speaking of overclocking, I'm not really afraid about it, but this is too fiddly and time-consuming, and I've seen my fair share of fellow computer scientists who roasted their hardware in the process. Moreover, to give you an idea, I'm the kind who generally has no problem deactivating anti-aliasing if it can help cooling. Maybe there's another reason too, I studied mathematics and computer science, both fields I was OK with (contrary to many computer scientists who have a problem with maths), but on the other hand I've always been bad at chemistry, and all those pastes, pads and other liquids that come along with the process of cooling, that reminds me too much of chemistry. Give me screws ! but no pastes, no liquids...

Now, on to my GPU ! I will come back once it has fai... ahem, I mean with the results.

 

Finger crossed for you! You'll come back once it has failed to fail! LOL
To preserve my hardaware, i use V-SYNC and with my 980m i must limit some game to 30FPS. For exemple Assassin's Creed Odyssey run at 32-40FPS depending on where i am and GPU is @99% so temp raise to around 80°C. If i enable a limit to 30FPS, the card will not work at max and stay at 73°C. The game remain the same! I need to found a good setting for FF15, this game is very very very demanding! So, all games need to have its own setting and need to be adjusted to find a good balance.
EDIT : there is a DOS version of NVFLASH. You'll need to disable UEFI boot.

 

Ah well, I'm sad to say it crashed again :

Code:

NVRM: GPU at PCI:0000:01:00: GPU-7868f4a9-abe7-e159-d3ba-a828745cc8ce
NVRM: Xid (PCI:0000:01:00): 13, pid=24604, Graphics SM Warp Exception on (GPC 1, TPC 0): Out Of Range Address
NVRM: Xid (PCI:0000:01:00): 13, pid=24604, Graphics SM Global Exception on (GPC 1, TPC 0): Physical Multiple Warp Errors
NVRM: Xid (PCI:0000:01:00): 13, pid=24604, Graphics Exception: ESR 0x50c648=0x34000e 0x50c650=0x4 0x50c644=0xd3eff2 0x50c64c=0x7f
NVRM: Xid (PCI:0000:01:00): 13, pid=24604, Graphics SM Warp Exception on (GPC 1, TPC 1): Out Of Range Address
NVRM: Xid (PCI:0000:01:00): 13, pid=24604, Graphics SM Global Exception on (GPC 1, TPC 1): Physical Multiple Warp Errors
NVRM: Xid (PCI:0000:01:00): 13, pid=24604, Graphics Exception: ESR 0x50ce48=0x36000e 0x50ce50=0x4 0x50ce44=0xd3eff2 0x50ce4c=0x7f
NVRM: Xid (PCI:0000:01:00): 13, pid=25033, Graphics Exception: ChID 0008, Class 0000b197, Offset 00001b0c, Data 1000f010
NVRM: Xid (PCI:0000:01:00): 62, pid=25033, a3a6(2198) 00000000 00000000
Asynchronous wait on fence NVIDIA:nvidia.prime:39ff timed out (hint:intel_atomic_commit_ready+0x0/0x54 [i915])

At first things were starting badly, with my Gnome desktop freezing as soon as the nvidia modules were loaded, but I waited for some time witout any Xid error so the driver seemed happy after all ! So I killed the display and started one anew, running 100% on the dGPU, and it worked. Then I tried glxgears (a small OpenGL program displaying spinning gears, that's often used to check graphic functionality on Linux), that was OK.

I started writing a victory message here, but before posting I thought I should try something a little more demanding, which I did with Gremlins vs Automatons, that's a card game, not very demanding but still more than what I did before. I stayed on the menu for one minute or so, then it froze and I rebooted before things became completely out of control. The system managed to reboot but there were tasks not responding and filesystems that couldn't be unmounted.

So the Xid error 62 (internal micro-controller halt) is back. I'll have to check what Xid 13 is... Maybe I will try starting the GPU with ECC mode activated (nvidia-smi -e 1). While it possibly degrades performance, maybe it would help determine where the problem is.

Of course the idea that something has been overheating comes to my mind. If it's the core, then that would mean the laptop's EC doesn't interpret the sensor readings correctly. I checked based on this assumption to see if hardware throttling was on, but it wasn't (however I checked only once). The displayed core temp was around 40°C most of the time, which to me seems consistent with a GPU that's not working much.

Now I'm considering a vBIOS flash, I doubt that's the problem, but there aren't many options left ! Could it be that I'm not applying enough contact with the VRM with the 2 mm pads ? After all, maybe 2 mm on the R22 inductors and 1.5 on R56 is too much and hinders contact with the nearby VRMs ?

 

With the inductors having any contact is fine, you just don't want the contact to be interfering with any other component.

 

I was thinking the same as @Meaker@Sager said. Try without any pads on the inductors, just 2mm on the VRM and 0.5 on the VRAM. What seems strange to me is the "out of range" errors.
I'll PM you today to send you my stock VBIOS if it still fail.

 

@Meaker@Sager @Shin Falcon I just reinstalled with 0.5 mm less on the inductors, now it shouldn't be more than what it was on my 880M hence I doubt it would pose problems with the VRM (still at 2 mm of course). Still didn't try yet...

There's something strange about that GPU crash, it's that it seems to kill the whole PCIe bus as well. I've noticed that shortly before a "GPU has fallen off the bus" error and a complete freeze, I get hardware error reports from the Wireless driver too. I had the idea to try with the wireless driver unloaded just in case, but it still crashed.

I've had that "fallen off the bus" error with my 880M in the past too, but I then knew it was entirely my fault because I had shut down the dGPU while the Nvidia drivers were still loaded, and then it wasn't freezing the whole system. Next time when I get the Xid error 62, I should try shutting down the GPU immediately to see if it prevents a freeze.

I can understand that the device has problems, but why should it pollute the whole bus, especially since the Nvidia driver reported a micro-controller halt ? Usually when a device fails, tasks using it can get stuck but the rest continues to work normally, not here though, the whole bus seems to be blocked, which gives a hard freeze and no more apparent activity, even when I'm on a text-only console that's managed at 100% by the iGPU.

I'm going to flash that vBIOS, and try to find Nvidia's Validation Suite and HW Field Diagnostic software (the latter runs on DOS). @Shin Falcon Concerning your vBIOS I remember you posted its version number, I suppose I can then get it from Techpowerup, but I can take it directly from you too if you think it's safer.

 

Hi, about my vBIOS, i didn't found it on Techpowerup.
You mention a PCIe kill, did you try without your wireless card? I doubt it'll help, if it was the case you'll have some errors about this. I didn't PM you yesterday since one of my friend told me he lost his father (Covid) and last week it was my wife's uncle (an AVC)... I'll not forget today!

 

Ah well, please accept my condolences ! I too know people who are potentially vulnerable to this virus (including my 95 y/o grandmother), but I've been luckier for now...

Concerning your BIOS, indeed I noticed it's not on Techpowerup. Speaking of which, there are many Clevo BIOSes there, in addition to yours I wouldn't know which one to use. Of course I should use an 8GB one, but concerning the clocks, iirc they're all the same. As a start, I think I will use the only validated Clevo BIOS.

And of course, it still crashes, within seconds like before. However, I did manage to prevent a complete system freeze by powering down the GPU after the Xid 62 error. Of course the driver complains with a "fallen off the bus" error, but then I can remove it (modprobe -r nvidia_uvm nvidia_drm), power it back on, and retry.

I also tried to keep spare MTRR memory-mapping registers this time in case the Nvidia driver would use it. Normally the systems keeps some spare by default but then I would not have optimal RAM coverage, hence I had a "no spare MTRR" policy (mtrr_spare_reg_nr=0 kernel parameter), but I reverted to the default (2 spare registers), which didn't change anything, still crashes. Anyway, I don't think modern GPUs need spare MTRRs any more nowadays, since their days of shared memory are long gone. As an example, my 880M didn't need them... but for now I will keep those registers spare, at least until that dGPU problem has been sorted out.

I'll try running nvidia-bug-report.sh after the Xid error, hoping that the freeze won't come before it has finished. I would have preferred the field diagnostics software, but it looks like Nvidia makes it available for GPU manufacturers only, the term "field" meaning here manufacturing field. Contrary to what I've seen some people say, it's not part of the CUDA toolkit which I installed.

Oh and maybe I should tell about something I noticed some time ago. Before I ran GPU-Z on my dGPU, I tried with Nvidia Inspector, and I was surprised by the vRAM clock displayed, that was 1598 MHz, the default clock being, heck, 2505 MHz. GPU-Z however, displayed something more consistent with regular clocks for a 980M, so I don't know if either Nvidia Inspector was wrong, or it has another meaning...

 

Hi @NovHak

Sorry i didn't forgot to send you my BIOS but just didn't had time yesterday. 95Y/o, if she's still feeling strong and well then keep her safe hope she'll be like this for still a long time
About your problem, it's still very strange... 1598 Mhz is indeed registered in the bios as P01 profile but GPU-Z may not show this speed only the max.
I was searching for a way for you to flash yout vBIOS. Can you boot in safe mode on Windows?
I asked this cause you'll need a modified version of NVFLASH to bypass the board ID check. I didn't have time to search if there is a version for DOS but since i see "x64" on the .exe, i'm pretty sure it'll not work under DOS.
If you can boot under safe mode, you can flash using CMD with administrator and use the -6 option to bypass the GPU MISMATCH caused by the difference of SUBVENDOR ID (remember the 1558 something i talked about some days ago?)
I did a test using this bios, and it did change my ID and working well. I choose this BIOS cause i don't need to edit my drivers anymore with this ID.
For now, just see if you can boot "stable" under safe mode.

 

Hi !

I understand you have other things to do, no worries about that, it's my problem first and foremost and not someone else's... and to be honest, sometimes I do other things too, including a bit of (non-performance) gaming, just to change my mind !

Concerning Windows, I have no problem going to safe mode, right now it's even configured with the dGPU disabled in the device manager, so I can boot into normal mode.

However, the Linux version of nvflash that's on Techpowerup has that -6 switch, so it looks like I can do everything on Linux. I already saved my current vBIOS (84:04:22:00:12) and compared it with the same version on Techpowerup just in case, and it's exactly the same file.

For what it's worth, here's my nvflash_linux -v output :

Code:

NVIDIA Firmware Update Utility (Version 5.414.0)
Simplified Version For OEM Only
Adapter: GeForce GTX 980M     (10DE,13D7,1558,7481) H:--:NRM  S:00,B:01,D:00,F:00

Identifying EEPROM...
EEPROM ID (EF,3013) : WBond W25X40A 2.7-3.6V 4096Kx1S, page
Reading adapter firmware image...
IFR Data Size         : 852 bytes
IFR CRC32             : CE15CE66
IFR Image Size        : 1024 bytes
IFR Image CRC32       : 7A49C7C4
IFR Subsystem ID      : 10DE-110B
Image Size            : 184320 bytes
Version               : 84.04.22.00.12
~CRC32                : C82C26F1
Image Hash            : FEE2E47B0A2C4516232250689E510E8B
OEM String            : NVIDIA
Vendor Name           : NVIDIA Corporation
Product Name          : GM204 Board
Product Revision      : Chip Rev
Device Name(s)        : GeForce GTX 980M
Board ID              : E906
PCI ID                : 10DE-13D7
Subsystem ID          : 10DE-110B
Hierarchy ID          : Normal Board
Chip SKU              : 700-0
Project               : 2754-0003
CDP                   : N/A
Build Date            : 09/03/14
Modification Date     : 09/05/14
UEFI Support          : Yes
UEFI Version          : 0x20005 (Jul 31 2014 @ 18745622 )
UEFI Variant Id       : 0x0000000000000006 ( GM2xx )
UEFI Signer(s)        : Microsoft Corporation UEFI CA 2011
InfoROM Version       : N/A
InfoROM Backup Exist  : NO
License Placeholder   : Absent
GPU Mode              : N/A
Sign-On Message       : GM204 P2754 SKU 003 VGA BIOS

Now my fingers are itching to perform some flashing, starting with the vBIOS you just mentioned, but I'll probably wait in case you would have a better advice !

EDIT : Finally I just flashed the only Clevo verified BIOS on Techpowerup (84.04.2F.00.79), but it still crashes promptly. I wanted to try the one you mentioned, but it says there's a board ID mismatch, which my utility apparently doesn't want to override (there seems to be an --overrideboard option, but it isn't recognised). I suppose now is time to go to Windows, since there are modified programs there. I will do that later today, unless maybe you advise me not to. For now I reflashed the GPU back to the previous, original BIOS (i.e. 84:04:22:00:12).

 

Ah nice! If you can go to safe mode, try this version of nvflash
I suggest you create a folder on the root of a USB drive then you can download it on linux and use it in windows. Don't forget to add your vBIOS!
You can lunch it via running CMD as administrator, going on the driver where the exe is, then "nvflash5xxx" (you may just rename it in nvflash.exe) -6 romname.rom
The -6 is important to bypass the ID mismatch.
You'll see some error, this is normal.
If you have an error about CERTIFICATE 2.0, then a driver is loaded. You should not see this error in safe mode.

BTW i just bricked my P150SM by flashing a bad BIOS on the mainboard. Seems like it try to read something from a USB drive but since i don't know which BIOS to put and what filename to use, i'll better buy a cheap CH341A programmer.

 

Fine, I'll try this and keep you informed

Bad luck having bricked your mainboard ! I remember I've read somewhere about an emergency procedure in such a case, as long as the BIOS boot block is OK. The problem indeed is to guess what the BIOS filename should be, and I remember seeing people who complained about Clevo changing the Amibios default names... However I've seen somewhere that there are programs that can analyse the BIOS and tell what the filename is, but it was a Windows program so I was too lazy to get it. I suppose you flashed the boot block along with the rest, hence the emergency flash procedure must be based on your bad BIOS' boot block. I will try to find that program and send you a link...

 

Thank for your help!
It's strange since i just flashed back the 1.03.05 stock BIOS. It flashed OK, then reboot normally. I press F2 to enter bios, load optimised then save and exit and... Black screen.
In case you didn't know, loading a bios file with Phoenix tool, you'll always have after the "working procedure" where it check the file, i pop up windows saying the recovery filename. And even with this and the good BIOS on a USB drive, it seems like reading, there is activity even on the hard drive and seems like to retry in a loop.
@Meaker@Sager if you have an idea, i know i may open a tread for my problem, but if you know how to complete a bios recovery on a P150SM, you'll be my heroe!

EDIT : I DID IT! After 4-5 hours searching a way to recover, i tried with my old 770m just in case : black screen. Just for the joke, i tried without dGPU and the joke turned bad : the screen displayed normally! Of course annoying beeping and shut down after 30s.
And here begin the impossible mission : flash a BIOS in less than 30s. Prepared my USB drive with Rufus and everything i need (freedos and of course the full bios package) and i cross finger on my hands and feet. System halted after flashing both BIOS, and it was reading the ME firmare to 34% when it shut down.
10 minutes laters, i'm back and everything work as before!
Now, it's time to sleep a bit
@NovHak hope you'll be as lucky as i was very soon!

 

Well that was risky!

 

Just to have it out of the way, you told about a Phoenix tool, but the P150SM is an American Megatrends (AMI) BIOS, not a Phoenix BIOS, or did I miss something ? At least that's what I see concerning the latest official 1.03.05 main BIOS, it uses afudos.exe, which is the flashing tool for Aptio BIOSes, hence AMI.

From what I've read, the procedure seems rather simple :

1. Put the BIOS file with the appropriate name (which is the most problematic to find) on a FAT-formatted USB stick. I've seen some say plain old FAT (FAT16) is better, and the formatted partition should be no more than 2GiB. I suppose that's the most conservative approach indeed, even though I'm not sure it's really necessary, and nowadays a huge FAT32 volume may be OK.
   
2. The laptop has to be powered off (power cord removed). And btw, after flashing the main BIOS, once the computer has shut down, the power cord shoud be removed for a few seconds (Clevo says 5s, Prema says 30s, I do 1 min lol)
3. Plug the power cord
   
4. Insert the USB stick
5. Power on
6. If everything is OK, the procedure should begin. It seems that it can take time to start, depending on the previous bad BIOS I suppose, but once it's done the computer shuts down.
7. Unplug the power cord
8. As said before, wait for some time
9. You should be OK, and can use your computer as before.

But how should the BIOS file be named ?

I've seen someone say that after putting an empty USB stick and powering on the laptop, a hidden file had been created with the name that should be used, and the same size as a BIOS file. Maybe you should try this. The default name is AMIBOOT.ROM, I've seen some say AMIBOOT.BIN, but as said before, it seems it's often changed by the laptop manufacturer, and there's some logic here after all, it helps making sure you're not mistakenly blind flashing with the BIOS for another machine... hence, if the empty USB stick method doesn't work, seeing how Clevo names their official BIOS files, I would try P15SM.ROM or P15SM.BIN.

Maybe it's the EC BIOS that's incompatible with current main BIOS ?

I've seen (but don't remember where) Prema saying for some flashes, that EC+BIOS have to be flashed in one go, or the machine will be bricked. That seems to be because some EC+BIOS combinations won't work. If that's the case, maybe you could try blind-flashing to the BIOS you were using before... but I don't know, really.

Last but not least, I've seen this thread about someone who bricked his P150SM, with Prema posting. They talk about an Fn+B procedure, but I've not been able to figure out what it is exactly.

EDIT : Just noticed you solved your problem, good news ! Now I have a theory : the problem was already solved, you didn't have to reflash

Concerning my own problem, I didn't boot Windows yet to do some more nvflash, but I will soon. I do have some hope, but frankly, I'm not overly optimistic !

 

No, really, you can open AMI BIOSES with Phoenix Tool, you can use it with Insyde BIOS too, i remember i used it to edit the BIOS of my old VAIO (2010 with i5 and radeon 5650m soldered).
I used it to extract the vBIOS and then overclock the card a little. it was a pain since the voltage was fixed to a lower value and so the GPU can't go very higher than what it was originally.
Since my P150SM is working again now i'll stop playing with BIOSES.

 

Yes, flashing again and again is risky ! On my GPU it's another matter of course, since it looks dead already... So I tried flashing with the 84.04.85.00.14 you were using, it went well, overriding the board ID successfully, but the GPU still crashes. I tried on Windows instead of Linux, it's less annoying now that I configured the boot manager to wait indefinitely so I have all the time I need to press F8 and go into safe mode, instead of having to wait for W10 to crash three to four times in a row before it finally decides to propose safe mode.

Btw, I discovered that the Num Lock led switching on during Windows' boot process seems to mean it has reached the Windows Boot Manager stage. Considering F8 works at this stage, maybe pressing F8 when Num Lock goes on would work, even if it's not configured to wait indefinitely.

I tried flashing another BIOS but this time is says GPU mismatch, because the PCI IDs differ :

Code:

NVIDIA Firmware Update Utility (Version 5.667.0)
Copyright (C) 1993-2020, NVIDIA Corporation. All rights reserved.


Checking for matches between display adapter(s) and image(s)...

Adapter: GeForce GTX 980M     (10DE,13D7,1558,7481) H:--:NRM  S:00,B:01,D:00,F:00


EEPROM ID (EF,3013) : WBond W25X40A 2.7-3.6V 4096Kx1S, page

WARNING: None of the firmware image compatible PCI Device ID's
match the PCI Device ID of the adapter.
  Adapter PCI Device ID:        13D7
  Firmware image PCI Device ID: 1617
WARNING: Firmware image PCI Subsystem ID (10DE.110B)
  does not match adapter PCI Subsystem ID (1558.7481).
WARNING: None of the firmware image compatible Board ID's
match the Board ID of the adapter.
  Adapter Board ID:        E901
  Firmware image Board ID: E923

NOTE: Exception caught.
Nothing changed!



ERROR: GPU mismatch

However it's still considered a 980M :

Code:

NVIDIA Firmware Update Utility (Version 5.667.0)
Copyright (C) 1993-2020, NVIDIA Corporation. All rights reserved.


Sign-On Message       : GM204 P2754 SKU 003 VGA BIOS
Build GUID            : Blank
IFR Subsystem ID      : 10DE-110B
Subsystem Vendor ID   : 0x10DE
Subsystem ID          : 0x110B
Version               : 84.04.88.00.1F
Image Hash            : 4165600033B25F7D5B39E157775D072A
Product Name          : GM204 Board
Device Name(s)        : GeForce GTX 980M
Board ID              : 0xE923
Vendor ID             : 0x10DE
Device ID             : 0x1617
Hierarchy ID          : Normal Board
Chip SKU              : 700-0
Project               : 2754-0003
Build Date            : 08/04/15
Modification Date     : 09/16/15
UEFI Version          : 0x2000C
UEFI Variant ID       : 0x0000000000000006 ( GM2xx )
UEFI Signer(s)        : Microsoft Corporation UEFI CA 2011
XUSB-FW Version ID    : N/A
XUSB-FW Build Time    : N/A
InfoROM Version       : N/A
InfoROM Backup        : Not Present
License Placeholder   : Not Present
GPU Mode              : N/A

Did you already try flashing when the PCI IDs are different ? I suppose that would require another flashing tool that's even more unlocked, but would it really be a good idea ?

Now my GPU is probably dead, but I was wondering, maybe if I install a specially crafted BIOS with reduced clocks, it would work in sort of a degraded mode ?

 

You could use a tuning program to lower the clocks like MSI afterburner in windows.

 

My preference goes to Nvidia Inspector, but anyway I can't since it crashes too fast once the driver is loaded, so I think I'll rather use the Maxwell II BIOS Tweaker to modify the BIOS my GPU came with, and flash with that.

EDIT : I've looked at a few tutorials but MBT still doesn't look overly clear to me.

My main goal now is to have the card working without crashing. So as a start I would like to have a vBIOS setup that's the least demanding possible (but with the card still working of course). Then I could try increasing some parameters and see what it does.

Considering it works for a few seconds, I suppose that some component reaches a threshold that makes it stop, hence my idea to prevent that threshold from being reached.

I intend to set the Max Table Clock in the Boost Table to the lowest possible value (1013 MHz), and set Boost, 3D & TDP base clocks to that value. (I'm using the 84.04.22.00.12 BIOS)

I suppose I could set those base clocks even lower without breaking anything ? (the lowest value being 405.5 MHz).

Maybe have a look at the voltage table too, lowering some values here and there could please the VRMs ? Though I suppose lowering the clocks by itself will logically lower power consumption...

I'm supposing much here...

 

There was a vbios editor at one point IIRC.

 

Maybe you're speaking of Nibitor. It seems very old though, and I doubt it would work with a Maxwell BIOS. Anyway I think now the point is made that the card isn't functional, and the following is more out of curiosity than anything : having a failsafe vBIOS, sort of. It could be interesting, for diagnosis purposes, to have a very minimal vBIOS that could work with a wide range of GPUs (of a given model that is, i.e. a wide range of GTX 980M in the present case), ensuring it can perform operations, though in a very degraded state.

Of course depending on the damage type, there are cases where no vBIOS will make the card run, but I'm thinking of those cases where it would run, such as possibly my case.

 

No, not nibitor a notebook community one.

 

Hello Everyone,

I am the OP of this post, and something really weird happenned today with the new GTX 980M that I have bought back in July:

I was playing Counter-Strike: Global Offensive and the screen just froze and the laptop became unresponsive. Then, I had to force shutdown. When I try to turn it on again, a black screen after the bios splash screen is present. The GPU it appears to do not boot at all after this.

I then managed to go into Safe Mode (This wasn't even working, also a black screen, I had to go to CMD to force the safe mode in minimal video), and Used DDU to uninstall the NVIDIA Driver.

The computer boots up again. It tries to install an old driver from 2017 via Windows Update (That worked nice before). The black screen is back after the 2017 Nvidia driver is installed and the system was rebooted.

Then, I manage to disable the GPU in Device Manager (Going into safe mode again) to use the computer, and it is working right now. But, when I go to GPU-Z, I have 0MB of RAM For the GTX 980M.


Is it possible that my gtx 980m fried? (Given this 0mb shown in GPU-Z) Or this might be just a software issue (Two days ago a new nvidia driver was installed and a new Windows Update was installed, but I don't know, I used DDU)

 

If you get boot into Windows then the GPU is not fried. But wrong vRAM size may be an indication of hardware problems.

 

EDIT: I do not get boot into Windows when the GPU is correctly installed (i.e.: Installed drivers and let the system reboot to complete installation). I only boot into Windows if I disable the GPU in device manager. I read that perhaps this can be a MOSFET problem? I've seen people being successful in changing these.

If I disable the GPU in device manager (and keep using the iGPU) The computer works fine.

If I enable the GTX 980M and let windows install the driver, it will ask me to reboot (normal procedure to finish the driver installation). Then, when it reboots it does not go the login screen (I have a black screen after BIOS)

However, at the BIOS the card is still correctly recognized das a GTX 980M.

Here is a print of the GPU-Z Showing 0mb. Is that because I disabled it in the device manager or the GPU itself is faulty?

https://ibb.co/frMFjk9

 

In that case maybe the card is defective or the GTX 980M you have might need a different vBIOS. I have GTX 1060 installed in my P170SM-A.

 

Thanks for the answer!

So, considering this, I was also thinking that during the episode of the computer freezing, the vBIOS got corrupted. Is that possible (I am really unexperienced on the subject)? Then, I was thinking about reflashing the GTX 980M. I have seen one link here at the forum that has one vBIOS version, but it appears to be an older version. However, it claims to be a CLEVO GTX 980M (My model).

1) Can one reflash the GPU even if its a previous version? (I know that one cannot to this to the BIOS of the PC)
2) Should I use nvflash for that?

If you could point me just the direction to do this, it would be greatly appreciated. However, I think that the GPU is faulty. I am just trying to eliminate any possible causes.

Many thanks!

 

It's complicated the flash a 980M. I used to flash my 970M a lot, but I forgot most of the steps.

 

I have seen a tutorial for both gtx 980m and gtx 970m also, at techinferno, using nvflash. Do you know if these work?

And can I reflash any GTX 980M vBIOS version, even if it is an older version than my current one?

 

I think you can. I mean, it would be problematic if you didn't have the Intel integrated video card, but if the 980M flash fails, you can still boot your machine using the Intel card and flash it to some other bios.

 

Possible for a bad vbios to do this yes.

 

So do you think that is Worth trying reflashing the vBIOS? Even if it is another version number, but same manufacturer (CLEVO)?

Should I use nvflash in safe mode to do that, correct? From what I seen from the tutorials in the internet

Thanks!

 

I think with yours it's likely the cars is just dead.

 

I reflashed the vBIOS today, it was pretty simple. The GPU is still showing 0MB at the "memore size" field at GPU-Z (It has been this way since the episode of the computer freezing). I agree, the GPU died, and probably it is one of the VRAMs.

I measured the resistance of the power supply of the GPU and is still measuring 2k ohms, so it does not appear that there is a short circuit (I have seen this in the "Broken GTX 980m" post here at the forum).

I am thinking about baking the GPU. At this point, I do not have anything to lose...

 

Note the Nvidia cards preprogram the ram they expect, they don't dynamically scan it on startup like the AMD cards do so if a module fails it will kill the card.

 

You refer to the modules being the VRAM chips or anything else?

What do you think about trying to baking the GPU? I mean, I have nothing to lose by now.

I tried:

Used DDU
Clean OS install
Change GPU vBIOS
Reseatting the GPU, Reapplying thermal paste and cleaning with isopropyl alcohol
Changed thermal pads


Is still showing 0mb.

On an interesting note, when I clean installed Windows 10, when I tried to install the NVIDIA Drivers, it always gets a BSOD with "DPC Watchdog violation". I am pretty sure that GPU-Z and "System information" showing 0MB/ N/A with respect to the vRAM indicates that one of the vRAM chips failed.

 

I had such a problem with a GTX 1080 TI. I had water spilt on it and it showed only 4096 MB of RAM versus its full size. I was never able to fix it.

 

That's sad (Given it is a 1080ti). I will try to bake my MXM GTX 980M This week, I have access to a lab oven so I do not put it inside my domestic oven. I will update this thread with this saga

 

Yeah a lot of passives round the chips that could cause issues even with replacing the modules themselves.

 

@Khenglish Reading the forum I have seen that you are literally an expert on the subject. Would you mind giving an opinion on my issue? This would be greatly appreciated!

Basically I was playing CSGO, Computer froze and I had to force it shutdown. When turned on again, Black Screen after BIOS Splash screen. Tried all of the steps mentioned in the reply above. In addition, if I disable the GTX 980M in device manager (By going to safe mode in order to do this), the computer boots normally. In GPU-Z, It shows 0mb in memory size and "Unknown" in memory frequency/bandwith.

I am thinking about reballing.

 

Sounds like broken solder. You could reball, but first I suggest trying a reflow with flux.

 

Thanks for the reply!

Sorry, I meant reflow instead reballing.

I do not have the equipment for reflow with flux. However, I do have access to a lab oven (An industrial stove). So I can regulate the temperature quite well and try the "oven trick": Go to 385°F preheat and using aluminium foil insert the GPU for about 8-10 mins.

However, I read some of your replies to posts in the forum and I have seen that you suggested in one of then to have 217°C as a Temp. Setpoint instead of 385°F (196°C). Is that the case?

I will try to reflow on friday when I have some free time. Any suggestions on the overall procedure are greatly appreciated.