In plasma 4 the widget had an option to show the utilization of each core/cpu separately, e.g. one could put a bar graph with 8 individual bars on the panel to see the system load. In plasma 5 it can only show a single bar of average cpu usage of the entire system, which is much less useful. Reproducible: Always
>which is much less useful. Why is seeing per core useful?
Personally, I'd recommend not to remove functionality just because you're not sure what the use case is... the functionality was there, and people used it, so it was useful to them in their own use cases even if they differ from yours :-) As luck would have it, just today I upgraded to plasma 5 (kubuntu 15.04), and already found a good example. Looking at the average cpu at some points during the day showed what appears to be a little bit of use (average ~15%), hardly something to see on the one-bar average graph, which would seem to be perfectly reasonable for a lively system. After a few hours I happened to use htop for a different purpose, and noticed one process was taking up 100% cpu (core), i.e. one thread was maxing out the core it was running on. Further investigation found it to be a busy wait - a serious bug in that software (reported, and hopefully fixed). On a graph with 8 bars (on my current machine - maybe my next one will have 16?), it's very different to see all cores being slightly active with a total average of 15%, than to see all cores idle and one core at 100%, with a total average of 15%. One is very normal, and the other means some very heavy single-threaded workload is going on, and if it's for too long, there may be something really wrong. On a one-bar graph they would be indistinguishable. The separate bars give additional and very important information about the system load, which is what this widget is made to show. It's perfectly valid, as in the plasma 4 version, to make this a configurable option, so every one can see his system load in the way most useful to them. (btw, in plasma 4 clicking on the widget opened the system monitor. It doesn't seem to do that in plasma 5. I personally prefer htop, but I'm sure others used this feature as well).
Just wanted to add - if you agree to restore this functionality (please do! :-) ), it doesn't necessarily have to look like the current bars with lots of space between them etc. It's ok if all cores are packed into a single graph with no spaces between the bars (like in previous widget), or even using thinner bars so as not to waste too much space - as long as the information is clearly visible. And lastly, the tooltip in plasma 4 widget also showed the dynamic running frequency of each core separately, which was also useful at times to see the load patterns.
FWIW, we didn't go round removing things, they're completely rewritten and some things got missed. In any case it's good to know "why" in order to come up with the best fix.
(In reply to Amichai Rothman from comment #2) > The separate bars give additional and very important information about the > system load, which is what this widget is made to show. > > It's perfectly valid, as in the plasma 4 version, to make this a > configurable option, so every one can see his system load in the way most > useful to them. > Could not agree more. The old per core display was very useful to show at a glance the general state of cpu/core utilisation. > (btw, in plasma 4 clicking on the widget opened the system monitor. It > doesn't seem to do that in plasma 5. I personally prefer htop, but I'm sure > others used this feature as well). Separate bug/feature request? That is something very useful I miss as well.
I agree with the bug reports arguments. Please implement the old functionality, it was tremendously useful.
Yes - The visual representation of the usage of each core is very useful, for example to see that you have limited a very CPU intensive process properly to cores 0-3 and left 4-7 free for other purposes (tasksel). Many of us are sitting on the LTS with kubuntu backports ppa until many of these minor nits get addressed. Thank you
One more thing that would be helpful, if you could add in frequency scaling usage, you would be one step ahead of the Plasma 4 widget. For example, if a core is running at half the frequency but loaded at 100%, it should display at 50% since it is actually only running at 50% capacity. The old widget would display the frequency when the mouse hovered over the bar graph but never properly used it in drawing the bar graph.
I'd also like to ask for restoring this functionality. It is especially useful for my terminal server, where lot of batch tasks are run. Actually I'm waiting with upgrade to plasma5 until this feature is available. Thank you!
Still an issue in plasma5-workspace v5.5.4 Not having the option to see/view a lockup/high-load of single cores makes this widget not useful on multicore devices -- especially important if running VMs, where specific cores are dedicated to specific tasks.
This should not be filed under wishlist but as a pretty big regression bug, since this functionality exists in previous versions but is now broken. Just mentioning this in case it helps in prioritization, so it doesn't get pushed down the todo list :-)
Sir - I would be overjoyed to have the original functionality back. But if you are rewriting it anyway, how about reflecting the frequency scaling in the calculation of the percentage displayed by the bar for each processor? So if you are running at 50% load at 50% of the max processor frequency, you would display 25%. The details shown in the fly-over pop-up box would be exactly the same. Why miss a chance to finally get the display as useful as possible?
I have two notebook and install same version of kde on gentoo. One can show individual core and the other can not. I don't know what is different between them. If anybody how to find the solution to the one which can not show individual core information. thanks
Following https://git.reviewboard.kde.org/r/125006/, this was added back 7 month ago. Right Mouse Button > Configure System Load Viewer - set Monitor type to Compact bar - and then check [x] CPUs separately It does work for me, although the visual appearance could be improved. But that is a different story. Can anyone confirm?
I will close as fixed meanwhile, if not, please reopen or create a new report with the specific issues you see :-)
I can confirm - this functionality has returned and is working well. You may close the issue.