Issues with nVidia GeForce 8400M: hardware failure or driver?

SuSE_Lamer · 07-24-2022, 03:09 PM

Hello,

I own laptop Fujitsu-Siemens Lifebook E8410 with nVidia GeForce 8400M GPU. This laptop has Mint 19 installed. I use it from time, last time - in April. It was OK.

A couple of days ago I started it again - and saw some artefacts on the screen, like dotted lines. After Linux booted, the resolution if display was 1024 x 768 (although the original physical resolution is 1680 x 1050). Additionally I got a hint that system ran without hardware video-acceleration.

I tried to change driver from open to proprietary - it lead to failed boot. I also tried to run Mint 20.2 with XFCE -same problem with resolution.

I am a bit surprised, because this laptop worked a couple of years without any issue. I researched a bit - this GPU seems to be free from "nVidia-bug" (like Thinkpads T61-series). Can anyone give me a hint about possible reasons? What I also saw - artefacts disappear after laptop started, there is no "freezing" (like it happens in the case of dead GPUs).

Regards,
Andrey

P.S. if necessary, I can translate messages from German into English.

Debian6to11 · 07-24-2022, 11:50 PM

Do the artifacts only appear in the BIOS screen? Mint seems clear of those.

Did you try a live distro to see that it is consistent?

mrmazda · 07-25-2022, 12:23 AM

This unlikely has anything to do with FOSS software. You can see G84 working fine here:

Code:

# inxi -SGaz
System:
  Kernel: 5.10.0-11-amd64 arch: x86_64 bits: 64 compiler: gcc v: 10.2.1
    parameters: ro root=LABEL=<filter> net.ifnames=0 ipv6.disable=1
    noresume mitigations=auto consoleblank=0 plymouth.enable=0 vga=791
  Desktop: Trinity v: R14.0.12 tk: Qt v: 3.5.0 info: kicker wm: Twin v: 3.0
    vt: 7 dm: TDM Distro: Debian GNU/Linux 11 (bullseye)
Graphics:
  Device-1: NVIDIA G84 [GeForce 8600 GT] vendor: XFX Pine driver: nouveau
    v: kernel non-free: 340.xx status: legacy (EOL, try --gpu) arch: Tesla
    process: 40-80nm built: 2006-13 pcie: gen: 1 speed: 2.5 GT/s lanes: 16
    ports: active: DVI-I-1,DVI-I-2 empty: none bus-ID: 01:00.0
    chip-ID: 10de:0402 class-ID: 0300
  Display: x11 server: X.Org v: 1.20.11 driver: X: loaded: modesetting
    unloaded: fbdev,vesa gpu: nouveau display-ID: :0 screens: 1
  Screen-1: 0 s-res: 1920x2280 s-dpi: 120 s-size: 406x482mm (15.98x18.98")
    s-diag: 630mm (24.81")
  Monitor-1: DVI-I-1 pos: primary,bottom model: NEC EA243WM
    serial: <filter> built: 2011 res: 1920x1200 hz: 60 dpi: 94 gamma: 1.2
    size: 519x324mm (20.43x12.76") diag: 612mm (24.1") ratio: 16:10 modes:
    max: 1920x1200 min: 640x480
  Monitor-2: DVI-I-2 pos: primary,top model: Samsung built: 2009
    res: 1920x1080 hz: 60 dpi: 305 gamma: 1.2 size: 160x90mm (6.3x3.54")
    diag: 184mm (7.2") ratio: 16:9 modes: max: 1920x1080 min: 720x400
  OpenGL: renderer: NV84 v: 3.3 Mesa 20.3.5 direct render: Yes

The dots on the BIOS screen indicate a probable video RAM problem, possibly RAM itself failing, or poor connection between GPU and motherboard.

SuSE_Lamer · 07-25-2022, 12:21 PM

Hello Debian6to11,

Quote:

Originally Posted by Debian6to11

Do the artifacts only appear in the BIOS screen? Mint seems clear of those.

Did you try a live distro to see that it is consistent?

I tried it already with Mint 20.3 XFCE, as I wrote - same. And you are right, artefacts appear only before Linux starts (BIOS or splash screen). It surprises me a bit, because I saw already failed T61 or Dell D630 with nVidia - one sees immediately that something is wrong.

I made a photo of inxi-output. I see that nVidia has no driver assigned - although they are installed. As I said, everything was OK in April, when I used it last time.

Regards,
Andrey

cesarbergara · 07-25-2022, 01:00 PM

Hi. ¿May be an automated update change the X configs or libraries?
Have a nice day.

Debian6to11 · 07-25-2022, 01:05 PM

It is not a driver issue. Something (probably Video/RAM related) during boot fails to work at full potential. The system seems to take care of it somehow after it boots.

I would be more worried that some day you might loose the screen.

Quote:

Originally Posted by SuSE_Lamer

I made a photo of inxi-output. I see that nVidia has no driver assigned - although they are installed.

You may not have installed them correctly. Check if they are installed, I cannot tell you how.

obobskivich · 07-25-2022, 01:46 PM

Looks like classic bumpgate artifacts (older GeForce GPUs, especially on mobile, are notorious for bad solder joints on the BGA package due to a rushed transition to RoHS compliance - this was well documented in Apple systems, and largely ignored elsewhere - it primarily impacts GeForce 7000 series chipsets, but isn't exclusive to them, because its more of a problem with manufacturing from that time period than a specific uarch), unless you know the machine got hit/dropped hard which may have damaged internal wiring between the display and GPU or similar. But as Debian6to11 speculates, it likely is able to 'work around' (but probably unintentionally) by not asking for acceleration and thus taking the load off the GPU (or thermal cycling is 'helping' the bad joints). Either way, it probably is time to consider a replacement - that machine has to be something like 15 years old at this point.

mrmazda · 07-25-2022, 03:13 PM

Quote:

Originally Posted by SuSE_Lamer

I made a photo of inxi-output. I see that nVidia has no driver assigned

That broken old inxi v3.0.32 doesn't have much to say about graphics. Recent versions do much better. Current is 3.3.19. If your inxi packager hasn't blocked auto-updating, its -U switch will update it. Of course, that won't solve the hardware issue evident in the BIOS screenshot.

SuSE_Lamer · 07-25-2022, 03:20 PM

Hello,

Quote:

Originally Posted by obobskivich

Looks like classic bumpgate artifacts (older GeForce GPUs, especially on mobile, are notorious for bad solder joints on the BGA package due to a rushed transition to RoHS compliance - this was well documented in Apple systems, and largely ignored elsewhere - it primarily impacts GeForce 7000 series chipsets, but isn't exclusive to them, because its more of a problem with manufacturing from that time period than a specific uarch), unless you know the machine got hit/dropped hard which may have damaged internal wiring between the display and GPU or similar. But as Debian6to11 speculates, it likely is able to 'work around' (but probably unintentionally) by not asking for acceleration and thus taking the load off the GPU (or thermal cycling is 'helping' the bad joints). Either way, it probably is time to consider a replacement - that machine has to be something like 15 years old at this point.

thanks for the great explanation. I suppose, Thinkpads of T61-series or Latitudes D620/630 suffered on the same issue.

I know, this laptop is not a newest, but it has T9500 and is sufficient for me. Try to find T9500 now :-) I will take of replacement, but I'm surprised that it happened very suddenly, the laptop was even not intensively used.

I made a photo of dmesg-output. I think, driver nouveau fails to load - see red text on photo.

Regards,
Andrey

obobskivich · 07-25-2022, 08:28 PM

As far as "sudden failure" its more a matter of it being noticed now, and if you aren't regularly running 3D games or something else stressful it may not have been noticed as soon. Without knowing a lot about this specific model: could you get another one of the same type and swap in the CPU from this one as an upgrade?

Also just for peace of mind: how does this do with a live image fully independent of your installed distro?

SuSE_Lamer · 07-26-2022, 04:08 PM

Hello,

Quote:

Originally Posted by obobskivich

As far as "sudden failure" its more a matter of it being noticed now, and if you aren't regularly running 3D games or something else stressful it may not have been noticed as soon. Without knowing a lot about this specific model: could you get another one of the same type and swap in the CPU from this one as an upgrade?

Also just for peace of mind: how does this do with a live image fully independent of your installed distro?

as I said already, I started live image of Mint 20.3 with XFCE - same issue (coloured lines on boot splash screen and XGA resolution when booted.

I'm just wondering what does the error message of nouveau driver mean and how I can reinstall and reconfigure driver and X server. Perhaps, I can fix that issue with resolution. But anyway, I think about swapping CPU from my main "working horse" Thinkpad R61 with T9500.

Just as info: Fujitsu Siemens E8410 is / was one of the upper-level business laptops. Its speciality is presence of all possible ports, really all. And very robust and well-built.

Regards,
Andrey

mrmazda · 07-26-2022, 06:57 PM

The difficulty in removing proprietary NVidia drivers is inversely proportional to the ease of installing them, often because easy installation dispenses with instructions required to reverse the process. All elements of the NVidia driver installation process must be eradicated for nouveau to be able to work.

business_kid · 07-27-2022, 08:17 AM

Isn't there an '/uninstall' (or similar) option on the Nvidia Installers?
There used to be, but that wasn't today or yesterday.

Debian6to11 · 07-27-2022, 09:57 AM

In case there is not an option to uninstall (or choose a different diver) from Mint's driver program, you may want to try this wikidebian page. Do that at your own risk.

https://wiki.debian.org/NvidiaGraphi...Uninstallation

SuSE_Lamer · 08-07-2022, 03:38 PM

Hello,

Quote:

Originally Posted by Debian6to11

In case there is not an option to uninstall (or choose a different diver) from Mint's driver program, you may want to try this wikidebian page. Do that at your own risk.

https://wiki.debian.org/NvidiaGraphi...Uninstallation

it is definitely not a driver issue. I found several posts, describing other laptops having this GPU and having same issues. It is a problem with GPU. So, I decided to sell this laptop for spare parts and not to invest more time.
Regards,
A.