CPU Idles at 31% - Never Below

[Penurious Penguin]

My CPU idles at 31% with a pretty constant temperature of 47c. I presently have CPU Frequency Scaling active and set to On Demand. The available spectrum ranges from 800MHz to 2.5GHz and aside from my i5 never indicating "Turbo Boost", it never scales below 800MHz either, regardless of processes. I remember the good old days when my processors could reach down to 0% when inactive and I'd like to know why it now cannot breach below 31 percent. I have uploaded an image of CPU Frequency Scaling Monitor showing where it's stuck at (31%) and the System Monitor showing slightly conflicting data, with the averages below 30 percent and nearer -20. http://tinyurl.com/7qjh9fh

Perhaps I am silly, but CPU-scaling should fully support the lowest possible frequency, especially for power-saving settings. For it to be stuck at the lowest of pre-set frequencies, especially as high as 800MHz, is simply ridiculous. I see no reason for the CPU to idle at 800MHz when far less than such is required.

[wayne128]

may be just type top in terminal and view what processes are taking the few top cpu% so that you have some clues?

[Penurious Penguin]

I knew this was coming and refrained from prophylactically defending myself against it. I know quite well how to use top, but find it silly when System Monitor is a mere click away and does not update itself at rates prohibiting coherent examination. Don't get me wrong, I love CLI and kneel humbly before it with true reverence, but I have been mildly perverted by certain features of convenience. Now that I've dignified myself, unless you require proof, there are simply NO processes consuming that much CPU. Actually, System Monitor itself is the largest consumer. In other words, you may take me rather literally and safely conclude that my CPU is idling at 31% which is far below what any single process is using, and often significantly less than the grand unified collective of festering bloat that wiggles beneath my glistening GUI. It may involve a tolerance for cognitive dissonance, but I assure you, something is strange, aside from me.

Thanks for the quick reply though, it really is a practical suggestion. Just keep in mind, some forms of [Li]nux may be becoming less practical.

[homerscousin]

I might be a newbee, but this might be a start. When I first installed Mint last month System Monitor was one of the first things I looked at to see how efficiently my old machine was running. I came from XP and I pretty much had Speedfan always running to monitor use and temps. When I first saw what S.M. was reporting I thought holy shite Linux is nowhere near as efficient as XP. Since then I installed Conky. What it shows is much closer to what I expected. I have Conky running right now. I opened S.M. and the processes tab and watched Conky's report as it was running. CPU use went up a good 20% just running S.M. I think it is wacky. I won't use it. Conky right now is using .5%. Try anything else other than S.M .

[HughT]

hi Penurious Penguin,
both wayne128 and homerscousin are right. The act of measuring your cpu load with the System Monitor increases the cpu load - known in physics as the observer effect. See my results below, not dissimilar to yours.

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
893 root 20 0 143m 51m 9868 S 1.7 5.2 2:45.91 Xorg
1649 root 20 0 16724 3744 2996 S 0.7 0.4 0:01.04 upowerd
491 messageb 20 0 3468 1732 848 S 0.3 0.2 0:01.29 dbus-daemon
1828 hugh 20 0 30712 10m 8076 S 0.3 1.1 0:04.71 metacity
2820 hugh 20 0 2636 1120 848 R 0.3 0.1 0:01.57 top

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
3005 hugh 20 0 175m 21m 16m S 28.2 2.2 0:08.47 gnome-system-mo
893 root 20 0 151m 59m 10m S 3.6 5.9 2:50.51 Xorg
1828 hugh 20 0 30712 10m 8076 S 0.3 1.1 0:05.06 metacity
2052 root 20 0 5568 720 456 S 0.3 0.1 0:00.26 udisks-daemon
2821 root 20 0 0 0 0 S 0.3 0.0 0:00.15 kworker/0:0
3010 hugh 20 0 2636 1124 848 R 0.3 0.1 0:00.10 top

The System Monitor uses 28.2%

[homerscousin]

One more little dagger for System Monitor. A couple weeks ago I had it open and was switching tabs, processes and the next one (performance?) the one with the graphs. I'd look at 1 tab for 5-10 seconds then the other and back and forth. The gnome system monitor process was strange. It would jump wildly from 4- 80%. I saw it once reported as 116%. OK, that should not be possible. I'm not a coder but I thought any % output would have to be defined as between 0 and 100. Again, I won't use it.

[Penurious Penguin]

For some reason I don't remember cpufreq monitor applet ever (at least not for me) being fixed on any particular number when set to OnDemand -- and it's been years!. I guess I got confused by taking it literally, especially after sensors displayed my core temperature at 52C when relatively idle. I foolishly thought the CPU could be scaled below 31%, and that the icon would indicate actual cpu-usage and not just the limit at which it can be scaled.

I guess since "solved" wouldn't quite fit this non-issue, I will remove it in a few days.

Thanks for the input folks, and pardon the sour tone; I think the "freeze"-bug got me off to a rough start.

Regarding System Monitor, it has it's place, but like many other Linux apps, it can be pretty bizarre at times. I guess the code was written by the Federal Reserve or something and set to display cpu margins instead of percent.