Ubuntu – Correct steps to install NVidia driver + CUDA after fresh install of Ubuntu Desktop 16.04.02 LTS

16.04cudadriversgraphicsnvidia

I've just downloaded Ubuntu Desktop 16.04.02 LTS and made a fresh install.
Step by step, how should I install the NVidia driver + CUDA?

From lspci output:

VGA compatible controller: NVIDIA Corporation GK106 **[GeForce GTX 660]** (rev a1)

I haven't tried to install the NVidia driver at all, yet. Apart from updating, I have only dowloaded Blender 2.78c from blender.org and manually extracted it to my home folder, and I have installed FlightGear 2017.1.2 from an additional repository.

With both applications, the desktop freezes permanently just a few seconds or a few minutes after startup, except for the mouse cursor. Maybe it's the Nouveau driver, I don't know and I'm not addressing that issue here, either, because I need CUDA anyway.)

It's not a UEFI machine => maybe more chances.

I went to the NVidia support website in order to see what version it is.

I also downloaded the .run NVidia file but did not launch it, instead I installed the corresponding package:

sudo apt install nvidia-375

The installation procedure finished with no apparent accident and I got the terminal prompt back.

Maybe I should have issued the command which is reported to be critical in various threads:

sudo nvidia-xconfig

Well, I didn't, not to risk messing up with the installation procedure in case it had been fixed by now to a fully automatic procedure.

First reboot => boot manager OK, then the LUKS pass phrase input field appeared but my input was not being displayed with bullet chars as usual. Ctrl-alt F1…F6: black screen. Ctrl-alt F7: black screen with my passphrase on the top line.

Second reboot => boot manager OK, then not even the LUKS passphrase input screen, just black screen, ctrl-alt F1…F7 => black screen.

I'm going to reinstall from scratch right now then give it another try issuing that further nvidia-xconfig command after installing the package.

Best Answer

(Addition to my earlier answer.)

I can now avoid booting through recovery mode.

The problem was only with the input of the hard disk encryption pass phrase. The workaround was to avoid the "quiet" boot, I found how to do it on this page:
Boot into the command line console in Ubuntu 16.04

The minimal change to /etc/default/grub was to replace

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash"

with

GRUB_CMDLINE_LINUX_DEFAULT="text"

then run

sudo update-grub

This gives small text (more info on the screen). You can uncomment GRUB_TERMINAL=console to have "normal" text mode (less memory used presumably).

GRUB_CMDLINE_LINUX="text" doesn't seem to be indispensable.

That web page also describes

how to boot to text mode and eventually start up the graphic mode afterwards
how to set the default mode back to graphic

Booting into text mode can be interesting to leave free the maximum possible amount of video memory, for instance for Blender to GPU render animations in scenes with high resolution HDR environments, of course after using graphic mode and CPU render to setup the scene and create the animation (possibly checking it with OpenGL render to speed up the process).

I have no solution for the corrupted windows frames after suspend-resume, yet. The nvidia driver I'm using is still the most recent, apparently.

EDIT: that is now solved.

A software upgrade applied 2017-vi-02 included a new NVidia driver, 375.66, and new LibCuda version. Automated installation was straightforward, Blender and FlightGear still work perfectly fine, and I can now suspend and resume without having to kill gnome-session or logout and in again.

I only see "garbage" - meaning corrupted graphics - during a couple of seconds on resume on the left of the screen where the auto-hide launcher would appear by moving the mouse cursor there, but the "garbage" disappears without having to go anything.

(I'm still not using the "quiet" boot so I can't tell if the keyboard input of the hard drive encryption pass phrase would now work fine with it.)

Err... sorry I'm a newbie in ask ubuntu: now how do I close this... "issue"? Should I "accept" any answer here? THANKS A LOT AGAIN FOR ALL YOUR REPLIES.

EDIT: I selected this post as the one which carries the information which was most useful to me.

Related Solutions

Ubuntu – What do you do when Nvidia-CUDA drivers break Ubuntu

The answer here copied from the wiki worked to fix it.

The question is, why do Ubuntu's official CUDA packages break everything? I filed a bug.

Ubuntu – Problem installing CUDA 8 on Ubuntu 16.04 – Missing proc/driver/nvidia folder

After a lot of effort for a new fella on Ubuntu I got it working!!!

@ubfan1 helped me when he said the command lshw -c video the hint I needed to know that I was not running on NVIDIA's driver.

After that I tried to install without success (the secure boot was off). I got this error: The system is running in low-graphics mode” (see this question). Therefore, I boot using the command **Alt+Ctrl+F1* and removed all files from NVIDIA using the command below.

sudo apt-get remove --purge nvidia-*

With that I could come back to my Desktop, where I could use the System Settings tool from Ubuntu to install the drivers (Software & Updates ->Addtional Drivers->Using NVIDIA binary -version 375.82 from nvidia-375 (open source). See the print:

I reboot the system and everything was installed!!! After that I tested the command ./deviceQuery from the bin inside the samples and the test PASS!

./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "GeForce GTX 1050"
  CUDA Driver Version / Runtime Version          8.0 / 8.0
  CUDA Capability Major/Minor version number:    6.1
  Total amount of global memory:                 4041 MBytes (4237426688 bytes)
  ( 5) Multiprocessors, (128) CUDA Cores/MP:     640 CUDA Cores
  GPU Max Clock rate:                            1493 MHz (1.49 GHz)
  Memory Clock rate:                             3504 Mhz
  Memory Bus Width:                              128-bit
  L2 Cache Size:                                 524288 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  2048
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 2 copy engine(s)
  Run time limit on kernels:                     Yes
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 1 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 8.0, CUDA Runtime Version = 8.0, NumDevs = 1, Device0 = GeForce GTX 1050
Result = PASS

Lesson learned

Before to install CUDA, make sure all drivers of NVIDIA are running OK! You may install them like @ubfan1 suggest in this link.

Execute the following commands to check if the installation is the default and it is running.

$ cat /proc/driver/nvidia/version
NVRM version: NVIDIA UNIX x86_64 Kernel Module  375.82  Wed Jul 19 21:16:49 PDT 2017
GCC version:  gcc version 5.4.0 20160609 (Ubuntu 5.4.0-6ubuntu1~16.04.4)

$ sudo lshw -c video
[sudo] password for marlosdamasceno: 
  *-display               
       description: 3D controller
       product: NVIDIA Corporation
       vendor: NVIDIA Corporation
       physical id: 0
       bus info: pci@0000:01:00.0
       version: a1
       width: 64 bits
       clock: 33MHz
       capabilities: pm msi pciexpress bus_master cap_list
       configuration: driver=nvidia latency=0
       resources: irq:321 memory:a3000000-a3ffffff memory:90000000-9fffffff memory:a0000000-a1ffffff ioport:4000(size=128)
  *-display
       description: VGA compatible controller
       product: Intel Corporation
       vendor: Intel Corporation
       physical id: 2
       bus info: pci@0000:00:02.0
       version: 04
       width: 64 bits
       clock: 33MHz
       capabilities: pciexpress msi pm vga_controller bus_master cap_list rom
       configuration: driver=i915 latency=0
       resources: irq:318 memory:a2000000-a2ffffff memory:b0000000-bfffffff ioport:5000(size=64) memory:c0000-dffff

$ lsmod | grep nvidia
nvidia_uvm            647168  0
nvidia_drm             45056  2
nvidia_modeset        790528  5 nvidia_drm
nvidia              12701696  85 nvidia_modeset,nvidia_uvm
drm_kms_helper        151552  2 i915,nvidia_drm
drm                   352256  6 i915,nvidia_drm,drm_kms_helper


$ nvidia-smi
Fri Sep  8 19:47:17 2017       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 375.82                 Driver Version: 375.82                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 1050    Off  | 0000:01:00.0     Off |                  N/A |
| N/A   49C    P0    N/A /  N/A |    536MiB /  4041MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID  Type  Process name                               Usage      |
|=============================================================================|
|    0       996    G   /usr/lib/xorg/Xorg                             271MiB |
|    0      1827    G   compiz                                         177MiB |
|    0      2351    G   ...el-token=FDDD25D3486FDA0AB5CD0952493279C6    86MiB |
|    0     14381    G   unity-control-center                             1MiB |
+-----------------------------------------------------------------------------+

Just to check the secure boot you can run.

$ mokutil --sb-state
SecureBoot enabled

Best Answer

Related Solutions

Ubuntu – What do you do when Nvidia-CUDA drivers break Ubuntu

Ubuntu – Problem installing CUDA 8 on Ubuntu 16.04 – Missing proc/driver/nvidia folder

Lesson learned

Related Question