What's wrong with a high refresh frequency?

(In reply to Nate Graham from comment #1)
> What's wrong with a high refresh frequency?

Currently desktop looks "hectic" if you have some sensors on it and I guess it needs more CPU time. I my opinion it would be great to have an setting "refresh time" for sensors. Update every two seconds would be sufficient - like 'watch' command.

Thanks for the explanation!

For reference:
in (the current) ksysguard one can set the update interval under View -> Tab properties.
I don't know if this also applied to the widgets of if there is a separate setting.

(In reply to Richard Llom from comment #4)
> For reference:
> in (the current) ksysguard one can set the update interval under View -> Tab
> properties.
> I don't know if this also applied to the widgets of if there is a separate
> setting.

It has no impact...

(In reply to dS810 from comment #5)
> (In reply to Richard Llom from comment #4)
> > For reference:
> > in (the current) ksysguard one can set the update interval under View -> Tab
> > properties.
> > I don't know if this also applied to the widgets of if there is a separate
> > setting.
> 
> It has no impact...

Note that you filled this bug under plasma-systemmonitor, the "new ksysguard". If you are using ksysguard, please fill bugs under:
https://bugs.kde.org/describecomponents.cgi?product=ksysguard


plasma-systemmonitor doesn't ship with plasma yet, you may find it as git package in your distros repository.

(In reply to Richard Llom from comment #6)
> (In reply to dS810 from comment #5)
> > (In reply to Richard Llom from comment #4)
> > > For reference:
> > > in (the current) ksysguard one can set the update interval under View -> Tab
> > > properties.
> > > I don't know if this also applied to the widgets of if there is a separate
> > > setting.
> > 
> > It has no impact...
> 
> Note that you filled this bug under plasma-systemmonitor, the "new
> ksysguard". If you are using ksysguard, please fill bugs under:
> https://bugs.kde.org/describecomponents.cgi?product=ksysguard
> 
> 
> plasma-systemmonitor doesn't ship with plasma yet, you may find it as git
> package in your distros repository.

I mean the "widget" is also named "system monitor" :/

(In reply to dS810 from comment #7)
> (In reply to Richard Llom from comment #6)
> > (In reply to dS810 from comment #5)
> > 
> > Note that you filled this bug under plasma-systemmonitor, the "new
> > ksysguard". If you are using ksysguard, please fill bugs under:
> > https://bugs.kde.org/describecomponents.cgi?product=ksysguard
> > 
> > 
> > plasma-systemmonitor doesn't ship with plasma yet, you may find it as git
> > package in your distros repository.
> 
> I mean the "widget" is also named "system monitor" :/

Yeah, I know its confusing...

(In reply to Richard Llom from comment #8)
> (In reply to dS810 from comment #7)
> > (In reply to Richard Llom from comment #6)
> > > (In reply to dS810 from comment #5)
> > > 
> > > Note that you filled this bug under plasma-systemmonitor, the "new
> > > ksysguard". If you are using ksysguard, please fill bugs under:
> > > https://bugs.kde.org/describecomponents.cgi?product=ksysguard
> > > 
> > > 
> > > plasma-systemmonitor doesn't ship with plasma yet, you may find it as git
> > > package in your distros repository.
> > 
> > I mean the "widget" is also named "system monitor" :/
> 
> Yeah, I know its confusing...

Can you move it?

(In reply to Nate Graham from comment #1)
> What's wrong with a high refresh frequency?

Hi, it’s me again with my four systems that I upgraded from Plasma 5.20 to 5.21 the other day. ;-)

On three of those I set up four or five system monitor applets for different data, such as CPU, disk load, network load and temperatures. After the upgrade, the update characteristics has changed considerably. The network applet updates once a second, all other twice a second. On the first impression, the visual offset each measurement takes ranges from very small to huge. After some observation I think I found the problem:

My CPU applet only has one line and it moves in small steps of a few pixels.
My temperature applet basically hurls across the screen, and it has shows four lines.
The network applet with two lines is inbetween.
It seems the graph only ever stores 50 values, shared among all graphs. The CPU applet with its single line shows 50 time steps, the network around 25 (50 columns / 2 graphs) and the temperature 18 (50 columns / 3 graphs).

(In reply to Frank Steinmetzger from comment #10)
> and the temperature 18 (50 columns / 3 graphs).

Checked again, they are in fact only 16. (16 * 3 = 48, which is where the 8 came from in my error).

Let's keep things separate, please. There's a known issue where sometimes the history amount for a chart is not properly respected, there is at least https://bugs.kde.org/show_bug.cgi?id=433064 to track that.

The request here is for adjusting the update rate of charts. I do have a plan for that, which has become possible now that we are running at a higher update rate. It will mean you can select an update rate that is a certain multiple of the current update rate, so you can have charts update at 500ms, 1000ms, 1500ms, etc.

Created attachment 136189 [details]
flamegraph of plasmashell while running the misbehaving widget

I believe I moved the bug successfully.

I agree with the usability issues discussed, it basically shows half a second of history, with maybe 2 samples at a time, which is a lot less useful than it used to be prior to plasma 1.21. I dug through the source and it appears to be a regression introduced in https://invent.kde.org/plasma/libksysguard/-/merge_requests/98, but I don't know enough about compiling and running KDE to try bisecting it to make sure.

For me, a bug in this widget, likely this one, has caused plasmashell to take 100% CPU on one core of a beefy processor and make the rest of plasmashell to be laggy. I've attached a flamegraph of where the CPU went.

I had the CPU widget in the bottom bar of my desktop, with 24 CPU sensors, one per hardware thread.

This was not happening continuously and I don't know if I can cause it to reproduce again.

*** Bug 433774 has been marked as a duplicate of this bug. ***

*** Bug 433695 has been marked as a duplicate of this bug. ***