NVIDIA - sure wish I rememberd what had been installed

[GreyGeek]

My Epson WiFi printer never gives me an option to print both sides when using Kubuntu. Sometimes it doesn't even want to print on one side.


Installed nvidia-prime nvidia-370 and everything looked fine except my fan would never stop running. I reverted back to 352 and no fan issue. I see that repository added some drivers in the 360's and I think I'll try upgrading a bit at a time. Maybe 370 is over clocking my GPU a bit.

When I booted my computer this morning at around 7 AM and installed Neon I continued to configure it until a little after eleven this morning when a noticed appeared that I had 5 minutes of battery power left! I had forgotten to plug in the power cord. For about 30 minutes of that time I installed Minecraft and played it. It gave a FPS of 160, but I could have set it higher. There was no significant change in temp that I could detect. I would use a cpu temp widget but the one I used for 16.04 won't work in Neon, even with lm-sensors installed.

Also, I have tried about every way I can think of to run the hplip duplex plugin. As I said before, when I fire it from the HP-Toolbox GUI it runs fine until it asks me for ROOT's password and won't accept mine. My first thought was to run kdesudo /user/bin/hp-toolbox but kdesudo is NOT installed and appears to depend on Plasma4 libs and such. Using sudo doesn't work because it aborts with the error message "do not run as root". The only thing left to do is give root a password, but I dislike doing that. Rock and a hard place.

[oshunluvr]

I just wanted to throw this in for Jerry and whoever else is playing about with your video drivers and wants to do a performance check: glxgears isn't really a valid speed tester. It is intended to test GLX functionality.

The default graphics performance tester is gtkperf

The speed difference GreyGeek experienced was from having vsync enabled prior to the reboot and not after - likely the updated drivers defaulted to not synced when before it was set to synced. I'm not saying there wasn't improvement, just that glxgears isn't the way to test it.

Example:

My system with vsync enabled:

stuart@office:~$ glxgears
Running synchronized to the vertical refresh. The framerate should be
approximately the same as the monitor refresh rate.
303 frames in 5.0 seconds = 60.499 FPS

Note the bold sentence. If you have vsync set, it will glxgears will always show about 60. Vsync is to prevent tearing.

stuart@office:~$ __GL_SYNC_TO_VBLANK=0 glxgears
68063 frames in 5.0 seconds = 13612.426 FPS

[GreyGeek]

My Epson WiFi printer never gives me an option to print both sides when using Kubuntu. Sometimes it doesn't even want to print on one side.


Installed nvidia-prime nvidia-370 and everything looked fine except my fan would never stop running. I reverted back to 352 and no fan issue. I see that repository added some drivers in the 360's and I think I'll try upgrading a bit at a time. Maybe 370 is over clocking my GPU a bit.

EDIT- EDIT:
I bit the bullet and
sudo su passwd root
and gave root a password.
I updated the duplex plugin on the hp-toolbox front page link and then
sudo su password -l root
to cancel it.

[GreyGeek]

I just wanted to throw this in for Jerry and whoever else is playing about with your video drivers and wants to do a performance check: glxgears isn't really a valid speed tester. It is intended to test GLX functionality.

The default graphics performance tester is gtkperf

The speed difference GreyGeek experienced was from having vsync enabled prior to the reboot and not after - likely the updated drivers defaulted to not synced when before it was set to synced. I'm not saying there wasn't improvement, just that glxgears isn't the way to test it.

Example:

My system with vsync enabled:



Note the bold sentence. If you have vsync set, it will glxgears will always show about 60. Vsync is to prevent tearing.

True. My vsync was set to "Auto". With nouveau the fps was the vsync rate. With the nividia driver it was not limited by the display vsync rate. And, the fps depends on the size of the windows displaying

:~$ glxgears
Running synchronized to the vertical refresh. The framerate should be
approximately the same as the monitor refresh rate.
58172 frames in 5.0 seconds = 11634.238 FPS
62801 frames in 5.0 seconds = 12560.192 FPS
62675 frames in 5.0 seconds = 12534.862 FPS
62238 frames in 5.0 seconds = 12447.465 FPS
... glxgears window size was dragged out to almost fill the screen ...
12202 frames in 5.0 seconds = 2440.391 FPS
11892 frames in 5.0 seconds = 2378.357 FPS
11893 frames in 5.0 seconds = 2378.429 FPS
11913 frames in 5.0 seconds = 2382.579 FPS

the gears:

[GreyGeek]

I just wanted to throw this in for Jerry and whoever else is playing about with your video drivers and wants to do a performance check: glxgears isn't really a valid speed tester. It is intended to test GLX functionality.

The default graphics performance tester is gtkperf

The speed difference GreyGeek experienced was from having vsync enabled prior to the reboot and not after - likely the updated drivers defaulted to not synced when before it was set to synced. I'm not saying there wasn't improvement, just that glxgears isn't the way to test it.

Example:

My system with vsync enabled:



Note the bold sentence. If you have vsync set, it will glxgears will always show about 60. Vsync is to prevent tearing.

I installed gtkperf but its output leaves a lot to be desired.
Using your command:

:~$ __GL_SYNC_TO_VBLANK=0 glxgears
47772 frames in 5.0 seconds = 9554.391 FPS
50508 frames in 5.0 seconds = 10101.433 FPS
50753 frames in 5.0 seconds = 10150.567 FPS
50724 frames in 5.0 seconds = 10144.626 FPS

[oshunluvr]

Sorry GG, my post wasn't clear, just run gtkperf

A GUI pops up with "Test All" as the default. Just click "Start"...

stuart@office:~$ gtkperf

(gtkperf:5918): Gtk-WARNING **: Unable to locate theme engine in module_path: "adwaita",

(gtkperf:5918): Gtk-WARNING **: Unable to locate theme engine in module_path: "adwaita",

(gtkperf:5918): Gtk-WARNING **: GtkSpinButton: setting an adjustment with non-zero page size is deprecated

(gtkperf:5918): Gtk-WARNING **: GtkSpinButton: setting an adjustment with non-zero page size is deprecated
GtkPerf 0.40 - Starting testing: Tue Nov 1 20:16:12 2016

GtkEntry - time: 0.01
GtkComboBox - time: 1.06
GtkComboBoxEntry - time: 0.86
GtkSpinButton - time: 0.09
GtkProgressBar - time: 0.15
GtkToggleButton - time: 0.16
GtkCheckButton - time: 0.07
GtkRadioButton - time: 0.11
GtkTextView - Add text - time: 0.29
GtkTextView - Scroll - time: 0.12
GtkDrawingArea - Lines - time: 0.35
GtkDrawingArea - Circles - time: 0.61
GtkDrawingArea - Text - time: 0.19
GtkDrawingArea - Pixbufs - time: 0.05
---
Total time: 4.13

Quitting..
stuart@office:~$

[GreyGeek]

Sorry GG, my post wasn't clear, just run gtkperf

A GUI pops up with "Test All" as the default. Just click "Start"...

I had no trouble running it. I just don't know the significance of what I am seeing. Where is the comparison or matching to an fps reference?

GtkPerf 0.40 - Starting testing: Tue Nov 1 21:23:27 2016

GtkEntry - time: 0.02
GtkComboBox - time: 0.82
GtkComboBoxEntry - time: 0.69
GtkSpinButton - time: 0.07
GtkProgressBar - time: 0.08
GtkToggleButton - time: 0.09
GtkCheckButton - time: 0.04
GtkRadioButton - time: 0.06
GtkTextView - Add text - time: 0.14
GtkTextView - Scroll - time: 0.00
GtkDrawingArea - Lines - time: 0.22
GtkDrawingArea - Circles - time: 0.38
GtkDrawingArea - Text - time: 0.11
GtkDrawingArea - Pixbufs - time: 0.03
---
Total time: 2.74

BTW, that time nearly doubles when the app is resized to full screen, but I still don't know what 2.74 seconds relates to.

[oshunluvr]

It's a benchmark. 2.74 seconds is how long it took your system to run the tests. It's only meaningful when compared to other gtkperf tests. Turn off desktop effects an see if it runs faster or run it on your other computers. You can see yours is running faster than mine. It will be interesting (maybe only to me ) how large of an improvement I see when I upgrade because I'll be using the same card.

The other run I did __GL_SYNC_TO_VBLANK=0 glxgears was to show that by turning off vsync, glxgears shows a totally different result. the __GL_SYNC_TO_VBLANK variable turns off vsync temporarily. gtkperf runs it's tests based on different functionality and isn't effected by vsync.

[oshunluvr]

Also, you can see by running gtkperf full screen that drawing circles take a huge hit in performance vs. the default window size.

There are other benchmark tools. glmark2 is supposed to benchmark your GLX performance but all I get is a segfault when I run it. It's either broken or not compatible with Neon. I downloaded a CPU benchmark tool called "truebench" and I'm going to attempt to run it before my upgrade and then again after. I'll post about that on my thread about my upgrade.

[urdrwho5]

Installed it and found that NCDU was quite useful.

Don't know if you're interested, but 'ncdu' is an ncurses-based disk usage analyzer that will allow you to drill down without changing dolphin views - although there's not a thing wrong with changing dolphin views Just another tool to have in the toolbox.

ncdu and glances are two command-line tools that I recommend to everyone