CNC Simulation vs Dry Run — Which One Actually Protects Your Machine
Before running a new CNC program on a real part, every machinist needs to verify it is safe. Two approaches dominate: the dry run (running the program on the machine without cutting) and CNC simulation (running the program in software). They are not equivalent. Understanding what each method catches — and what each misses — determines which is appropriate for your situation.
What Is a Dry Run?
A dry run executes the NC program on the actual machine with cutting disabled. Common dry run methods:
Air cutting — run the program with the tool above the workpiece (Z raised by a safe amount) and spindle off, watching the machine move through the program at reduced speed.
Single block mode — execute one block at a time, watching each move before approving the next.
Feed rate override — run the program at 5–10% feed rate with a finger on the feed hold button, ready to stop if the tool approaches something unexpected.
Machine lock — some controllers have a "machine lock" mode that executes the program and updates the position display without actually moving the axes.
What a Dry Run Catches
- Obvious positioning errors — the tool goes to the wrong side of the part, or rapids to a location clearly outside the workpiece
- Over-travel alarms — the program tries to move beyond machine limits
- Incorrect work offset — the reference position is wrong and the program runs in empty space or inside the fixture
- Program syntax errors that stop execution — missing block numbers the controller requires, unrecognized codes
What a Dry Run Misses
Collisions at cutting depth. In an air cut with Z raised, the tool never enters the workpiece. A collision between the tool holder and a clamp at Z-50mm is invisible because the tool is never at Z-50mm.
Spindle-off behavior. With the spindle stopped and the feed rate at 5%, the machine dynamics are completely different. A rapid move that triggers an alarm at full speed may pass silently at 5% feed override.
Tool length errors. If G43 H2 is active but H1 was intended, the tool height is wrong. During an air cut the discrepancy is invisible because you're not measuring against the actual part surface.
First-part crashes. A dry run that looks correct at 5% speed with Z raised may still crash on the first part if the operator forgets to remove the Z offset, change the feed rate override back to 100%, or re-enable the spindle.
Subtle depth errors. A Z depth that is off by 0.5mm is invisible during an air cut and only shows up as a dimension error on the finished part — or a broken tool on the next operation.
Time cost. A thorough dry run on a complex program takes as long as the actual machining cycle. For a 45-minute program, that's 45 minutes of machine time spent not producing parts.
What Is CNC Simulation?
CNC simulation executes the NC program in software on a virtual machine model. The simulation interprets the actual G-code — the same file that runs on the controller — and shows the tool path, material removal, and machine motion in 3D.
A full machine simulation (as opposed to a toolpath viewer) includes:
- The complete machine structure — column, spindle, table, tool changer
- The tool assembly — tool, holder, collet
- The fixture and workholding — vise, clamps, step blocks
- The workpiece — stock material
- Controller-specific G-code interpretation
What Simulation Catches That Dry Runs Miss
Collisions at actual cutting depth. The simulation runs the tool to its actual programmed Z depth inside the 3D machine model. A collision between the tool holder and a clamp at Z-50mm shows up immediately in the simulation — before the machine moves at all.
Tool length errors. The simulation applies the tool length offsets to the virtual tool, making depth discrepancies visible.
Fixture collisions at rapid speed. Simulation runs at full programmed speed (in the 3D model), catching clearance problems that a slow-speed dry run would miss.
Post-processor errors. A syntax error that produces incorrect controller behavior shows up in the simulation's G-code interpreter, not just as a machine alarm.
Material removal errors. Simulation shows the finished part geometry — gouges, excess stock, and dimensional errors are visible before any material is cut.
What Dry Runs Catch That Simulation Misses
Real machine dynamics — resonance, vibration, actual acceleration behavior. Simulation models kinematics, not dynamics.
Setup errors — if the fixture is positioned wrong or the work offset was entered incorrectly at the machine, the simulation (which uses the correct offset) won't catch it.
Operator errors during execution — pressing the wrong button, wrong tool loaded, wrong offset entered. These are execution errors, not program errors.
The Right Combination
Simulation and dry run are not alternatives — they verify different things and work best together:
| Simulation | Dry Run | |
|---|---|---|
| Program logic and syntax | ✅ | Partial |
| Collisions at cutting depth | ✅ | ❌ |
| Tool length errors | ✅ | ❌ |
| Material removal | ✅ | ❌ |
| Machine dynamics | ❌ | ✅ |
| Setup/offset errors | ❌ | ✅ |
| Machine time required | Zero | Full cycle time |
| Catches first-part crash risk | ✅ | Partial |
Recommended workflow for new programs:
- Simulate in Eureka3X — verify program logic, collisions, depths, tool paths
- First-part dry run in single-block mode — verify setup, offset, operator execution
- Run at full speed — confident the program is correct
This approach catches program errors in simulation (where they cost nothing) and setup errors in dry run (where the machine is moving but not cutting), minimizing both crash risk and machine downtime.
The Cost Argument
A machine crash costs €5,000–€50,000 in spindle and fixture repair, plus days of downtime. A broken tap in a critical part costs hours of EDM extraction and potentially the part.
Eureka3X costs €9/month. A single crash pays for decades of simulation.
The question is not whether to simulate — it's whether to simulate before or after the crash.
Try Eureka3X free for 30 days →