Geforce Go 7950 GTX stability problems

[Naked]

Hi,

I have a Rock laptop with a Geforce Go 7950 GTX GPU and Intel Core 2 Duo with 2.6.18-2 debian kernel and I'm running into a lot of problems with it.

- The left side of the screen has a constant pixel flicker noise zone for an inch or so.
- X freezes every now and then for like 30 seconds.
- X sometimes randomly crashes during usage.

These problems are severe enough and happening often enough that I cannot use these drivers.

The problems occur with driver versions:

- 1.0.8776
- 1.0.9625
- 1.0.9742

I have tried the pci=nommconf option and idle=poll kernel boot options to no avail.

(PS. these problems look a lot similar to http://www.nvnews.net/vbulletin/showthread.php?t=80630 - and on that note, there's a Phoenix TrustedCore(tm) NB BIOS version 1.00.02 in this machine - and there's no note of driver updates available from the laptop manufacturer.)

[netllama]

I see a few WAIT warnings in your X log:
(WW) NVIDIA(0): WAIT (0, 7, 0x8000, 0x00006844, 0x00006844)
(WW) NVIDIA(0): WAIT (0, 7, 0x8000, 0x000007e0, 0x000007e0)

Those can be triggered by driver, kernel or BIOS bugs.

I have a few questions:
0) Are you able to capture the corruption in a screenshot? Does it persist at any mode, or just 1920x1200?
1) I'm not clear on your comment regarding the BIOS. Are you using the latest?
2) Does the instability persist with an official Debian kernel as well?

Thanks,
Lonni

[Naked]

Originally Posted by netllama

I have a few questions:
0) Are you able to capture the corruption in a screenshot? Does it persist at any mode, or just 1920x1200?

Yes, I was able to capture a screenshot. Taking the screenshot hung the machine for quite a while and the screenshot has more black lines than the screen normally has even.

The problem persists in other modes as well.

Originally Posted by netllama

1) I'm not clear on your comment regarding the BIOS. Are you using the latest?

Yes - there are no bios updates available from the manufacturer at all - so I'm using the BIOS that the machine arrived with.

Originally Posted by netllama

2) Does the instability persist with an official Debian kernel as well?

Yes. Atleast with a nvidia-kernel module compiled with module-assistant. I am unable to test the debian distribution nvidia-kernel-module because the nvidia kernel module in the distribution at the moment is for 2.6.18-1-686 where as the kernel there is 2.6.18-2-686.

[netllama]

I'd like to see your screenshot.

When I had asked whether the problem reproduces with an official debian kernel, I was referring to the kernel, not Debian's repackaging of the nvidia X driver. You can always install the official nvidia driver package with any kernel.

Thanks,
Lonni

[Naked]

Originally Posted by netllama

I'd like to see your screenshot.

I attached it to the earlier message now. I was cropping it so it fits withing the size limits imposed on attachments here - that is just the left side of the whole screenshot - but the rest of the screenshot has nothing new.

Originally Posted by netllama

When I had asked whether the problem reproduces with an official debian kernel, I was referring to the kernel, not Debian's repackaging of the nvidia X driver. You can always install the official nvidia driver package with any kernel.

It does appear on official 2.6.19-rc6. I have not tested on the latest 2.6.18 - if you want that it takes me a while to compile it.

Edit: Oh sorry, I misunderstood your last message - yes it does not work on 2.6.18-2-686 debian official kernel with the nvidia kernel module compiled via module-assistant.

[netllama]

Have you tried using the NVreg_Mobile nvidia kernel module parameter (see the README for more details)?

[Naked]

Originally Posted by netllama

Have you tried using the NVreg_Mobile nvidia kernel module parameter (see the README for more details)?

I tried it - tried modprobing the module with "NVreg_SoftEDIDs=0 NVreg_Mobile=3". I also tried all choices from 1-5 and none made any difference (well, number 2 caused a crash, but I'm not sure if it was any different really, could've been chance).

[Naked]

As an additional data point - if the nvidia driver has messed up the display in X, when switching to the console, the console also will have problems. There are odd noises in characters and general display corruption. This persists even after shutting X down and removing the nvidia kernel module.

So whatever the nvidia driver does when X is started leaves the display adapter in a state which causes display corruption.

[netllama]

I tracked down a notebook with a GeForce Go 7950 GTX, and didn't run into any corruption while X was running. This problem appears to be specific to your environment/notebook. Unfortunately, I do not have any Rock laptops here to test with.

With respect to the instability, if you could isolate it down to specific applications or reproduction steps, I could look into that further.

Thanks,
Lonni

[Naked]

Originally Posted by netllama

I tracked down a notebook with a GeForce Go 7950 GTX, and didn't run into any corruption while X was running. This problem appears to be specific to your environment/notebook. Unfortunately, I do not have any Rock laptops here to test with.

Thanks for looking into it.

Originally Posted by netllama

With respect to the instability, if you could isolate it down to specific applications or reproduction steps, I could look into that further.

Actually, I now have some clue to what might be going on.

I have windows installed on this laptop as well - and I tried some driver versions on the Windows side. Obviously, none of the nVidia publicly provided drivers would accept to install on a Geforce Go 7950 GTX laptop - but I used the modded .inf files from laptopvideo2go.com to force them to install.

All of the windows drivers I tried exhibited the same display corruption as the linux drivers did - except it was much worse, I couldn't even get to the desktop most of the time. But the drivers from the manufacturer work smoothly.

Which brings me to think that there is something special in the drivers I get from the manufacturer that make them work with the graphics card in this laptop. The drivers say to be of version 86.14 and they are available here: http://www.mobile-support.net/drivers/ms10514.zip

Now, if I can find out what's special in those drivers and replicate that on the linux side - I think I could make this work.

Any help would obviously be greatly appreciated.

[Naked]

Okay, it seems I was a bit off.

The drivers that are for Geforce Go 7 series (84.63) work fine after being forced to install on the Geforce Go 7950 GTX (by modifying the inf file).

So whatever is making the Geforce Go 7 series drivers special from the desktop drivers might be the cause.

[Naked]

Found it!

[nv_SoftwareDeviceSettings]
HKR,, PerfLevelSrc, %REG_DWORD%, 0x2222

If that key is missing - all the symptoms I described above happen on windows - and worse. If that is present, then any version of the driver seems to be working smoothly on windows.

So now I only need to figure out how to specify that, or something equivalent, to the linux driver!