m0154@tnc.UUCP (GUY GARNETT) (04/05/91)
"You will find that the truths we cling to depend greatly on our own point of view" - Obi-Wan Kenobi I checked the "system overhead" on my Amiga 1000 last night. The machine configuration is A1000 w/ 2,5Mb RAM, 3 floppies, Workbench loaded on RAD:, FaccII, QMouse, and ARexx installed. First, I used the Monitor utility on the Commodore v1.3 Extras disk. It showed about 25% system usage with nothing happening. Moving the mouse caused a small but detectable increase. Clicking on various windows and calling up the menus made the utilization go up to about 60% or so. Doing just about anything on the workbench or in the CLI made it max out (100% used) for a short time. Hmmm ... I never knew there was that much overhead in the system ... So I got out ASDG's SysMon (which came with FaccII) and tried the experiment again. With the system completely idle (no disks in drives, no mouse or keyboard operations) the thing showed no usage at all. None. Moving the mouse around made a small flicker on the graph (representing less thant 1/2 of 1% load). With disks in all three drives, I proceeded to actually use the machine to do real work. When I paused in the middle of what I was doing, I noticed (again with no keyboard or mouse activity) that SysMon showed about 2%-5% usage. Moving the mouse around made no noticable difference. Clicking on various windows, and calling up the menus increased the load to about 25% or so. Typing a text file to the CLI caused to load to go up to 100% for the duration of the operation. Very Interesting ... My conclusion is that the program used to measure the overhead is probably more significant that the overhead itself. I think that the results from SysMon more accurately reflect the actual CPU usage of the system (somehow, it figures that ASDG would do things right; I've never been dissapointed with their products), while the Commodore supplied program seems to use about 20%-25% of the CPU just to measure and display system utilization. If this is correct, the overhead for a copletely quiescent system is on the order of 1%. Even with some activity (disks in drives, mouse movement, disk cache programs) the overhead is around 5%. My system has a 7.14Mhz 68010 installed; adjust all numbers up by about 10% for 68000 based systems (let's say 2% and 6%; in any case, still significantly less thant 10%). For a 68020, 68030, or 68040 based Amiga, the operating system overhead is probably lost in the measurement noise (I would estimate it at less that 1% in all cases). Wildstar