startx fails: NVIDIA(0): Failed to initialize DMA.

Graphics / Linux Linux
1

Howdy,

I’ve Gigabyte GTX 650 Ti, nvidia 310.19 drivers, xorg-server 1.13.0 and kernel 3.6.8. When I run startx it starts off but fails somewhere with: “NVIDIA(0): Failed to initialize DMA.” in /var/log/Xorg.0.log. lsmod | grep nvidia lists the module is loaded.

Any thoughts what could I try to fix this? I tried 304.xx drivers and they just failed with a segfault.

// ville

2

Howdy,

Here is my nvidia-bug-report.log.gz attached. It has a superfluous .png extension in the filename because otherwise this board prevents the attachment.

// ville

3

Anything, is this thing on?

// ville

4

You should blacklist nouveau driver and vesafb too .

5

Nouveau module is blacklisted and vesafb turned off by passing ‘video=vesafb:off’ to kernel at boot. lsmod lists nvidia’s module as loaded.

// ville

6

Does nouveau actually work on this system? Did any version of the NVIDIA driver ever work? It sounds like there is a system-level problem that is preventing the driver from communicating with the GPU. Do you have a different GPU you could try, or can you try moving it from one PCIe slot to another?

7

Noveau doesn’t work with the card. As far as I am aware they don’t even support the chipset at all*. When I tried to boot with noveau all I got was a corrupted screen of white with random specs of bright colors.

I have an older card, GeForce 7600 GS, that runs fine on the same system with the 304.xx and older. 310.xx of course dropped support for that generation. I only tried the card in the one PCIe x16 slot and it didn’t occur to me to try it in a PCIe x1 slot.

Unfortunately I can’t right now test it in a PCIe 1x slot, because I am trying to get the store to take the card back and they have it. Apparently they tested it under Windows 7 and concluded it to work. Doesn’t help me much though.

// ville
*) http://nouveau.freedesktop.org/wiki/CodeNames - doesn’t list GK106

8

That’s pretty strange. I did notice one interesting-looking warning in your kernel log:

Does passing the pci=use_crs option to the kernel help?

You might also want to check with your motherboard vendor to see if an updated system BIOS is available.

9

I managed to get a different motherboard to try with similar results. I did find out that certain combinations of my RAM modules caused this behaviour, so I probably have a faulty one.

// ville