We've all seen the 'CPU load', be it in an overclocking software context, or even just simply Windows task manager or Linux's top. But what is it actually a measure of? CPUs are digital devices (a switch is on or off), but the scales of CPU load are analog (run from 0-100 in steps). My theories are:
With memory, the percentage is much easy to determine, as it's a percentage measure of capacity, but where CPU is a measure of work done/load, it could be calculated in many ways.
The list is just my theories, I'm not CPU engineer! Does anyone actually know the answer here?
- How many of the registers are full, and how full
- How many 'lost cycles' there are, based on latencies, probably as a moving average
- How full the L1/L2/L3 cache is
- A measure of percentage of IPS, as a defined maximum for a given CPU
With memory, the percentage is much easy to determine, as it's a percentage measure of capacity, but where CPU is a measure of work done/load, it could be calculated in many ways.
The list is just my theories, I'm not CPU engineer! Does anyone actually know the answer here?








