Calculate overall system load combining CPU, I/O wait and runnable process counts.
The three load average figures Linux reports (1, 5 and 15-minute exponentially-weighted moving averages of the run-queue length) each smooth over a different window, so comparing them to each other reveals a trend that no single number shows: 1-min running noticeably higher than 5/15-min signals load is currently rising, while the reverse signals a recent spike subsiding. Weighting them (0.5/0.3/0.2, favoring recency) into one combined figure gives a single trend-aware number for dashboards or alerting thresholds, without discarding the longer-window context entirely.
Weighted combined load
combined_load = load_1min × 0.5 + load_5min × 0.3 + load_15min × 0.2
Because they're independently-decaying exponential moving averages, not simple windowed averages — a short, sharp spike shows up strongly in the 1-minute figure almost immediately, but takes much longer to meaningfully move the 15-minute figure, which is exactly what makes comparing them useful for spotting whether load is currently rising or falling.
It counts the average number of processes in the run queue (runnable, waiting for CPU) plus processes in uninterruptible sleep (typically waiting on disk I/O) — not just CPU-runnable tasks. This is why load average can be high even when CPUs look idle, if many processes are blocked on slow storage.
Roughly, yes — load average equal to core count generally means the CPUs are fully utilized with essentially no queuing. Load average significantly above core count indicates tasks are waiting for a turn on the CPU (or for I/O), and load average well below it indicates spare capacity.