Diagnostics

The Diagnostics panel is collapsed by default. It reports on the hardware, the solver worker and the current run, and gives you a few recovery buttons.

Rows

Row	Meaning
GPU	Device name.
Compute	The CUDA compute capability, or the backend name for non-CUDA devices.
Memory	Free and total video memory, for example 4.8 GB free of 6.0 GB.
This scene	The estimated GPU memory the current scene needs, with a status of Safe, Warning or Critical.
Worker	Connected with the round-trip latency in milliseconds, or Starts on first run.

The estimate comes from the Blaze grid resolution, the Hydro or Atom particle count, or the Impact body count, compared with free memory.

When the status is Critical, the main panel also shows a box with Lower Resolution to Fit. See Out of GPU memory for the thresholds and what it changes for each solver.

Solver statistics

While a run is active, extra rows appear for the solver:

Solver	Rows
Blaze	Grid (cells per axis and pressure iterations)
Hydro	Particles, with the whitewater count
Impact	Bonds (and how many are broken) and Debris
Any	Backend, Substeps (marked adaptive with Auto Substeps) and Dynamic colliders

Buttons

Button	What it does
Refresh icon (Run System Doctor)	Re-reads the GPU, pings the worker, re-estimates the scene’s memory and re-audits the environment, then reports a summary line.
Restart Worker	Safely ends hung worker processes, resets sockets and starts a fresh worker.
Check Python	Searches your system for a Python with the GPU runtime installed and sets it in the preferences.

See Cannot connect to the worker and Out of GPU memory.