Cloud solver · What changed

CoolSim cloud solver changelog

Every change to how the cloud solver meshes, solves and reports your jobs, newest first, with a page for each release. A job runs on the release that was current when it started.
Ask about a change Desktop app changelog

Release 2026.10.23.16

Can change results

Improvements to solve startup for rooms with perforated tiles

Before the first iteration, the solver prepares every perforated tile in the room. We fixed an issue where this preparation could take a long time in large rooms with many tiles. Tiles are now prepared much faster, so these jobs reach their first iteration sooner. Each tile keeps the same open area and resistance. Tile airflow can differ slightly from the release before. Rooms without perforated tiles run exactly as they did on the release before.

Improvements to airflow checks in CRAC parameter studies

Before meshing, a CRAC Parameter Variation study checks every cooling-unit airflow in its table to catch a mistyped value. We fixed an issue where a valid airflow could be refused when the model’s own airflow for that unit was a placeholder. A value is now refused only when it is far above both the unit’s typical airflow and the table’s typical airflow. Studies that already ran are not affected.

Release 2026.10.23.15

Improvements to cooling-unit supply control

Each cooling unit’s supply air temperature is adjusted step by step as the solve runs. We fixed an issue where a model in which every cooling unit is set by temperature difference could stop when the solve started. Those models now run. Models with at least one cooling unit set by cooling capacity run exactly as they did on the release before.

Release 2026.10.23.13

Can change results

Improvements to air density in external-flow models

An external-flow model now takes its reference air density from the outdoor temperature you set, rather than from a fixed value. We fixed an issue where, on a site warmer or cooler than that fixed value, the outdoor air could behave as if it were rising or sinking on its own, which made some sites hard to solve and could affect the airflow around the building. External-flow results can differ from earlier runs, most on hot sites with light wind. Indoor models run exactly as they did on the release before.

An airflow check for CRAC parameter studies

Before meshing, a CRAC Parameter Variation study now checks each cooling-unit airflow in its table and stops when a value is far above what that unit normally carries, which usually means a mistyped value. The study stops before it uses any solve time.

Improvements to temperature ranges in result images

A temperature range you set for the result images is now read the right way round when its top value is entered first, so the legend runs from cold to warm. A range with no width is set aside and the images use the automatic range.

Improvements to pressure legends

Pressure images of a plane where the pressure barely changes now show the plane’s actual values in the legend.

Release 2026.10.23.10

Can change results

Improvements to fan-curve and cooling-capacity control

A cooling unit defined by a fan curve now moves toward its operating point gradually as the solve runs, and a unit in cooling-capacity mode adjusts its supply temperature in smaller steps. We fixed an issue where rooms cooled by fan-curve units in cooling-capacity mode could fail to converge, and an issue where a room already below its set point could be cooled harder. These models now solve more reliably. Results for rooms with fan-curve or cooling-capacity units can differ from earlier runs.

Improvements to recovering unstable solves

When a solve becomes unstable late in the run, it is now started again once with the other solver method before the job stops. If the solve had already met the convergence check several times in a row before it became unstable, that converged solution is delivered rather than lost. Solves that stay stable run exactly as they did on the release before.

Earlier checks for duct models

A model with a duct that has no diffuser is now stopped before meshing, rather than running and failing later in the job.

Release 2026.10.23.7

Improvements to how a finished run completes

In rare cases the solver could stop responding just after a run had finished, which held the job and could end it as a failure even though its results were complete. CoolSim now recognizes this and completes the job with its results. Results themselves are unchanged, and runs that finish normally are not affected.

Release 2026.10.23.2

Improvements to VFD cooling-unit control

A cooling unit under VFD control adjusts its fan speed as the solve runs to hold its setpoint. We improved how the solver makes those adjustments, so models with VFD-controlled units solve more reliably. Models without VFD control run exactly as they did on the release before.

Release 2026.10.23.1

Can change results

A more complete heat-balance check

Before a result is delivered, the solver checks that the heat leaving the room matches the heat put into it. That check now counts all the heat in the room, including PDUs, UPSs, switchboards and other powered equipment, not just racks. Models that carry much of their heat in power equipment are now judged on their full heat load.

Release 2026.10.23

Can change results

Updates to unit temperatures in the report

The report now reads each unit’s inlet and outlet temperatures on the unit’s own faces, and tells inlet from outlet by the direction air flows through it. This covers PDUs, UPSs, chillers and other units that give off heat, and outlet temperatures now line up with each unit’s heat load and airflow. The solve itself is unchanged, but the unit temperatures in the report, and the picture of unit inlet temperatures, can differ from earlier runs.

Release 2026.09.30.4

Can change results

Improvements to ceiling grilles and cable cutouts

October 8, 2026: the result report went back to an earlier design, so jobs from that day on no longer get the rebuilt report or the updated energy section described below. The other changes in this release still apply. Ask us if you need to know which report a job’s results are in.

A ceiling grille or cable cutout drawn slightly off its ceiling or floor is now set in place before meshing, so it carries the airflow you gave it. Airflow through these openings, and the temperatures it feeds, can differ from earlier runs.

Improvements to how the mesh aligns objects

Can change results

When the mesher aligns objects on shared planes, each object now keeps its full thickness, and a model that is meshed again without gaps keeps the shapes you drew. Meshes can differ from earlier runs.

More consistent rounding in the report

Can change results

Numbers in the report are rounded more consistently, so some values can read slightly differently than in earlier reports.

Improvements for models without cooling units

Models with no cooling units now go through solver setup reliably.

Near-miss acceptance for models without rack heat

The near-miss acceptance added on September 22 now also applies to models with no rack heat.

Updates to the report’s energy section

Can change results

The energy section now accounts for fan power and for heat through the walls more accurately. On models that enter a cooling COP, the PUE in the report can differ from earlier runs. On October 8, 2026, it was withdrawn: see the note under this release’s headline.

A rebuilt result report

The result report is rebuilt. Its 3D view opens in your browser in place of the ANSYS viewer, with see-through baffles and an air-speed scale that covers all the air in the room. On October 8, 2026, it was withdrawn: see the note under this release’s headline.

Release 2026.09.29.1

Can change results

Improvements to meshing

October 8, 2026: the report went back to an earlier design, so jobs from that day on no longer get the Summary Report change described below. The other changes in this release still apply. Ask us if you need to know which report a job’s results are in.

The mesh stage is rebuilt. It manages how cells grow and how gaps between objects are meshed more carefully, retries a model that fails to mesh using what it learned from the failure, and stops before solving a mesh that is too poor to give a trustworthy result. Meshes, run times and results can differ from earlier runs.

Updates to how the Summary Report rates racks

Can change results

The Summary Report now colors and counts each rack by its warmest and coldest inlet points rather than the average across the inlet, so a rack with a local hot spot at its inlet is flagged. On October 8, 2026, it was withdrawn: see the note under this release’s headline.

More precise temperature conversion in the report

Can change results

Temperatures are converted from kelvin more precisely, so some values in the rack tables and some image legend limits can read slightly differently than in earlier reports.

Additional checks before a result is delivered

Can change results

Before a result is delivered, the solver now checks that air temperatures and speeds are within a physically plausible range. A solve outside that range is stopped rather than delivered.

Release 2026.09.26.3

Can change results

Improvements to fan-curve cooling units

October 8, 2026: the report went back to an earlier design, so jobs from that day on no longer get the redesigned Summary Report described below. The other changes in this release still apply. Ask us if you need to know which report a job’s results are in.

Cooling units defined by a fan curve now account for air density more accurately when they remove heat from the room. Results for rooms cooled by these units can differ from earlier runs.

A redesigned Summary Report

Page 1 of the Summary Report now opens with the verdict, four headline figures and the key findings, so the result reads at a glance. On October 8, 2026, it was withdrawn: see the note under this release’s headline.

Improvements to solver stability

The temperature solution is stabilized from the first iteration, which helps more models converge.

Improvements for models with blocks and power units

Models that list a plain block ahead of a PDU or HDU now go through solver setup reliably.

Release 2026.09.22.3

Can change results

Independent unit control in cooling-capacity mode

In cooling-capacity mode, each cooling unit now runs its own controller instead of sharing one with every other unit, so each unit is adjusted independently. Supply temperatures, and the room, can differ from earlier runs.

Improvements to racks with chimneys

Can change results

When every rack in a row that exhausts through the rear and top has a chimney, the exhaust is now routed up the chimneys. Airflow and temperatures around those racks can differ from earlier runs.

Improvements for rear-and-top exhaust racks

Models with racks that exhaust through both the rear and the top now go through model setup reliably.

Release 2026.09.22

Can change results

Updates to convergence acceptance

A solve that comes very close to meeting the convergence check, with its heat balanced and its temperatures in a believable range, is now delivered rather than refused.

Release 2026.09.13.2

Can change results

A more thorough convergence check

A solve now keeps iterating while it is still settling, and a result is accepted only when it has converged and, in a model with rack heat, its heat balances. Some runs take longer, and some that would have been delivered before are now refused.

Improvements to meshing small gaps

Can change results

The minimum gap between objects now reaches the mesher more accurately, which gives better cells between objects that sit close together. Meshes and results can differ from earlier runs.
This changelog starts with the release of September 13, 2026. Changes that leave your mesh, results, report and email the same are not listed. For the desktop app, see the CoolSim 5.4 changelog.
Read and watch

Results different from an earlier run?

Send us the job number. We’ll tell you which release it ran on, which change it is, and what it means for your model.

Ask about a change