Measuring Simulation Performance in Northstar at Quadrupled Speed

On 23 September 2026, performance measurements were conducted in CS2 for Northstar between 08:29 and 12:57 UTC at a population of half a million residents and quadrupled game speed. The evaluation examined the active simulation rather than a paused or loading state. Rather than making the city appear faster by simulating less, the objective was to sustain additional residents at unchanged quadrupled speed while limiting compute work per simulation frame.

Process Work and Throughput

The evaluation placed two metrics side by side: CPU seconds per 1,000 frames as a measure of process work, and processed frames per real-time second as throughput, alongside actual population changes. Seven sufficiently long measurement windows classified as valid were recorded for that day. An initial window of around 50 seconds was discarded due to its short duration and was not included in the dataset.

Four examples show why a single peak figure reveals little:

Process work per frame varied visibly, whereas measured throughput remained within a narrower range across these examples. These differences do not constitute a controlled before-and-after effect of any specific construction measure, as rendering, camera angles, diagnostic routines, and city states differed between windows. The population difference represents a net figure within each respective window rather than a count of departures.

Infrastructure as a Hypothesis

During the investigation, an alternative cross-connection of approximately 250 meters was constructed. Vehicles were subsequently observed using the new segments, but the northern bottleneck persisted. Later, six stops and a new bus line were added. Following a fleet expansion, seven buses were assigned and active; at 13:00 UTC, 119 passengers were inside the vehicles while 935 waited at the stops. The line was in active use, but the queue had not cleared. Neither a reduction in car trips nor a CPU gain is proven by these snapshot values.

Current Status and Open Points

A new road can open an alternate route without easing outflow at the subsequent intersection, just as more buses can operate while demand outpaces capacity. A lower CPU figure can stem from a different scene or load. A robust comparison will require longer, uniform measurement windows at matching speeds, similar camera placement, and documented interventions, while taking population trends and warning levels into account.