Intermittent graphics corruption on text boot

Graphics / Linux Linux
1

About one in 20 boot cycles on Ubuntu 18.04 with a GTX 1070, I get graphics issues on the text console. Really ridiculous corruption in some cases:
https://photos.app.goo.gl/RUVQZfZKfBkpirEP7

Setting nvidia-drm.modeset=1 in /etc/default/grub improves things, so that the system doesn’t hang or have obvious visible corruption. But still 1 in ~50 cycles xorg won’t come up and in each case I see nvidia-smi showing 100% GPU usage (where in normal boot cycles it’s at 0%). The xorg error I get is:
[ 21.625] (EE) AddScreen/ScreenInit failed for driver 0

At the moment the system has a more severe manifestation, with nvidia-smi hanging.

Here’s the bug report .gz:
https://drive.google.com/file/d/0B8R6VSvTZHZWME1DR25hN2Z2U2lSSEZRZkxLaXVRU2JVLTc0/view?usp=drivesdk

2

looks like this:
[url]https://devtalk.nvidia.com/default/topic/1026107/linux/-solved-xid-62-fixeable-/post/5222318/#5222318[/url]

3

Thanks for your reply. That looks like it was a hardware issue, whereas I’ve seen my issue in multiple identical GTX 1070 cards.

4

Most of the time you’re getting an XID 44, sometimes an XID 31 and once also a

NVRM: RmInitAdapter failed! (0x26:0xffff:1123)

You could try with a 410/415 driver but I’d rather suspect an unstable power supply, sometimes failing when the gpu throttles up on driver init.