I don’t get it. First, I have Screenlets running under Python because they look cool and I can see the displays from across the room. Then, when top tells me that Python is hogging CPU to the tune of thirty-plus percent when I have nothing else running except for Gnome itself (I even shut down Pidgin), I close all Screenlets, kill Python, and switch to the “lightweight” option…Conky. Conky reports 25-27% usage…why? According to top this time, it’s either Xorg or nvidia-settings, and seems to be mostly nvidia-settings (polling GPU core temp, presumably, but that seems awfully frequent??) Okay…so let’s kill Conky and restart all screenlets except for the sensors screenlet that is set to display the temp output from nvidia-settings; the rest just use output from sensors, and of course, Python. That drops CPU usage to 9%, so an “aha” moment; it’s got to be nvidia-settings. Now, what happens if we start the GPU core temp screenlet? Answer? Nothing. CPU usage stays between 9% and 10%. Start Pidgin…no change (once it’s loaded). Now, start Firefox 3.5. No big change; CPU usage goes 10-11% and stays there. Nautilus causes a significant jump, but that’s because Nautilus can be a pig, especially with very large directories and several mounted partitions, all of which I have, and I know that, so I don’t run it unless I’m actually using it. This doesn’t make any goddamned sense. There’s nothing in my conkyrc that should be causing high CPU usage. If anything, Screenlets should be more resource-intensive because they run atop Python, and have a “pretty” graphic interface, not just a text interface like Conky. I even tried removing the GPU information block from conkyrc, but that didn’t change a thing–still stayed at 27% (also worked like shit, since wget was refused over and over for external IP information, but that’s beside the point). So, it would seem that if I let Screenlets start on login as they do now, but stop them all, and restart them, I get the CPU usage report I would expect to get, but if I don’t restart them, I’ll get an extra 10-20% usage, and that’s backed up by the output from Conky (and by top in all cases). WTF?