2

I have some issues to get two monitors to display from the GPU. Or even one, for longer than 5 seconds :)

This is my system:

  • supermicro X10SRL-F
  • intel xeon E5 1620 v4
  • MSI NVIDIA GeForce GTX 1060 6GB 6GT OC V1
  • 32 gb ram
  • etc.

And I installed Ubuntu 16.04 (including third-party packages).

I have one monitor connected to the VGA of my motherboard, and one with the displayport to the GPU.

Upon booting, both screens show the purple "Ubuntu ..." screen but then only the VGA remains on and shows me the login screen.

I've tried all kinds of cable-switching and Ubuntu tips I found online. It unfortunately doesn't work.

  • I first added the recommended ppa´s to the sources.list (https://launchpad.net/~graphics-drivers/+archive/ubuntu/ppa)
  • then I ran: sudo apt-get install nvidia-387
  • updated + upgraded, and rebooted (repeated this last step to be sure)
  • no changes.. (both screens again shortly show the first Ubuntu screen upon booting (the GPU works) but then only the VGA stays on)

I read about disabling secure boot from the BIOS, but I can't seem to find this anywhere in the BIOS. I changed some PCI things, mostly setting things from disabled to enabled or auto, but this seemed not really relevant (could look up details if needed, but the effects are unchanged).

I would want to be able to first use them in two-display setting, and of later with some changes to do some machine learning. But just having two screens to work on in the first place would be a thrill :)

Any help is very welcome, thanks!

(disclaimer: I was unsure where to post, so this is a cross-post from https://stackoverflow.com/questions/48341020/nvidia-gpu-supermicro-motherbord-ubuntu-16-04-not-working)

Improve this question

1 Answer 1

Reset to default
1

You cannot run both the integrated XEON graphics and the discrete NVIDIA card at the same time. You should switch to the NVIDIA in bios and run all of the screens from NVIDIA.

You could, however, use another computer to control the machine Learning computer in a headless situation, like via SSH...

Improve this answer
8
  • Thanks - so I am in theory not able to do machine learning on the GPU while having one screen via the onboard VGA? That would be a shame.
    – nick88
    Commented Jan 19, 2018 at 15:36
  • Of course you can still do machine Learning on the card when it is being used for graphic, or make it a headless machine like in my updated answer.
    – Nothingismagick
    Commented Jan 19, 2018 at 15:46
  • Hm, thanks, I will look into this, didn't think of ssh yet.
    – nick88
    Commented Jan 19, 2018 at 16:00
  • Just a quick question about changing VGA priority in the BIOS: what happens if I change this to offboard: imgur.com/a/aDYOL (screenshot from a picture online, but it looks the same) - will I risk not being able to even attach a screen if I can't get the GPU to work?
    – nick88
    Commented Jan 19, 2018 at 17:09
  • lscpi does give me this: imgur.com/a/W1tm5
    – nick88
    Commented Jan 19, 2018 at 17:13

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .