Northstar on 2 October: growth, supply and a new efficiency yardstick

Northstar's last reading of 2 October showed about 1.66 million residents, below the day's peak at 19:07 UTC. More than two million residents have not been reached.

The report covers 2 October by Berlin local time, from 1 October 22:00 UTC to 2 October 22:00 UTC. All times below are UTC.

The day in numbers

The daily file holds 66 population points and one documented gap in the grid. Missing values stay unknown and are not interpolated. The series contains restart and loading phases, so it is not one uninterrupted simulation run.

The cost of growth

The housing subsidy widened the evening before proved expensive. In the morning the building agent recorded a negative monthly balance, and the operating log attributes about 100 million of the spending to housing subsidies. After a human request to stop the outflow and run the city more robustly, taxes were reset, and the next budget reading showed a positive monthly balance. That is a change in the running balance, not a refund of the money already spent. An isolated tax effect cannot be determined, because declines, restarts and other measures lay between the compared population levels.

Another costly trial was withdrawn. In the afternoon, paid external services were switched on for a short test. The monthly balance turned from positive to clearly negative, and garbage warnings fell only slightly. The services were switched off again.

New residential areas: utilities first

Several new residential areas were created. W4 and W5 together had about 5,500 zoned residential cells, and W7 and W8 followed with further zoning in the same range. These figures describe zoning, not residents or completed homes.

The first areas showed problems in the order of development. In W7 and W8, power warnings were documented. After two substations and underground high-voltage cables were added, the local checks went to zero. Later, warnings in a south-eastern industrial area were no longer found after a new substation and cable route. Such zero-warning checks are local results, not a guarantee for every district.

Emergency and funeral services were also missing near the new districts. A hospital, clinics, crematoria, schools and other services were added. A later operations check reported ambulance warnings falling sharply; several service measures and ongoing operation lay in between, so the report does not credit one building alone.

For later areas, such as W9 and W10, services and power connections were planned before zoning. Failures remain part of the account: one residential area had been placed by mistake on a resource-rich tile and, because homes had already grown, was not demolished. In W8 a badly planned road across a lake had to be partly removed. In W12 a cable first ended without a correct connection; the substation was realigned and connected, and the later local check found no warnings at 41 homes there.

Funerals and garbage: more vehicles are not enough

Funeral service stayed a problem. Cemeteries were replaced, crematoria added, and garages and extra ovens built. An evaluation error in the garage tool led to repeated expansions: the log lists 38 garages at eleven crematoria instead of one controlled addition. The evaluation was corrected, but funeral warnings had not yet fallen. In the evening the inventory distinguished between the fleet on hand and vehicles actually in use. The building agent treated the large number of deceased citizens as an open question of supply and counting, not as a proven vehicle shortage. A lasting clearance of the backlog is not shown in the day's material.

For garbage collection, distance and a lack of large building plots kept recurring. A recycling plant was built in the east, and six office buildings gave way to it. Other sites failed on remaining overlaps despite a young office building having been demolished beforehand. After 20:00 another plant followed in the central east. Checking its effect remains an open task.

Freight rail and mail

A new freight station in the east was linked to an external connection with double track, and a freight route was then active. A follow-up check found an empty train and material warnings at almost as many businesses as before the build. The businesses still received ore and timber over roads. An active route therefore does not yet show a shift in freight flows or resolved material supply.

For mail, full post offices without active vehicles were examined in the evening. The building agent diagnosed a spatially mismatched sorting chain. An existing sorting plant was expanded and four new ones were built in the centre, east, south and west. One planned plant did not fit even after two industrial businesses were demolished. Later sites received their own short access roads, and a candidate affecting homes was dropped. The city then had nine sorting plants. Whether mail backlogs and the related satisfaction factor improve lastingly had not been checked by the end of the day.

From 18:00, the optimisation target changed

Until the afternoon, core masks and repeated restart draws were examined. An aborted comparison series recorded lower CPU load with reduced core release, but also clearly less simulation throughput. It was not a complete, controlled long-term comparison. Lower CPU load with a slower simulation does not meet the efficiency goal.

At 18:00 the human target was restated: more than two million residents with less demanding processing inside the simulation. The earlier percentage limits now serve as guide values, not as a hard success rule. Targeted re-draws by restart were ruled out, and the CPU controller was set to all logical processors. Memory-related restarts remain as crash protection.

The new assessment is meant to relate CPU time to the simulation work done and to the population. Short loading phases, pauses or population loss must not produce apparent success, and process, world and speed transitions must stay separate. The expert review also distinguishes fewer vehicles from fewer pathfinding searches: a bundled public transport trip is not automatically a correspondingly smaller number of individual trip plans. The size of the expected relief is a hypothesis, not a measured result.

First measures followed. A duplicate bus line was deleted, and an additional depot track let a metro line run more trains. A measurement note shows temporarily more pathfinding work after the line change and extensive queries, a transition effect. Later, traffic lights at selected junctions were removed in a first citywide series, with four-arm junctions treated separately. A follow-up comparison at the same game hour was not yet available, so no congestion or CPU success is claimed.

Move-in without a re-draw

In the evening, move-in stalled again. Instead of re-drawing, the building agent examined tax incentives and an entry district near a connection. The residential tax was raised to ten and later twelve percent. An area zoned as low-rent housing had produced no residential buildings after the documented short wait; demand there was zero, and it stayed a reserve area.

The diagnosis of a limited house-hunting budget comes from internal code analysis. The empty district is consistent with it but does not prove the full cause of the stalled move-in. The later fall in population and the growing stock of households wanting to move in remain unsolved. The goal of not letting the population fall is a target, not a confirmed state.

Status and open points

The performance file holds about 2,600 raw readings and 33 documented gaps, a count that includes the edges of the day; the counts of process and world identifiers are not a reliable number of restarts, and CPU differences across such boundaries are not normal performance windows. Open: the effect of the new recycling plants, the mail backlog, the traffic-light follow-up comparison and the cause of the stalled move-in. Project page: CS2.