General Parameters
The General Parameters rollout controls the fundamental properties of your simulation domain — in other words, it defines the core settings of the FumeFX Grid.
FumeFX GPU features powerful NVIDIA® CUDA acceleration that delivers, on average, 2× to 5× faster simulation performance compared to CPU*. The FumeFX GPU solver is designed to overcome typical GPU memory limitations: if the simulation exceeds available GPU memory, FumeFX automatically switches to CPU mode, ensuring even the most complex scenes can be completed without interruption.
Despite the speed boost, FumeFX GPU faithfully preserves the realistic fire and smoke behavior the software is known for. While many sparse fluid solvers lose detail or motion quality, FumeFX GPU maintains both - providing fast simulations without compromising high-end visual accuracy.
*Measured on a system with Ryzen 7900X and RTX 4090.
Simulation Area
Spacing - This parameter directly affects the detail level in your simulation. Type in or use the spinner to set the spacing of the voxel-based grid. The chosen value defines voxel size; smaller voxel size allows for more voxels in simulation, which produces greater detail.
Width, Length, Height - These three values set the volumetric dimensions of the FumeFX Grid, thereby establishing the boundaries for simulation within the scene.
Blocking Sides - On each grid axis, you can choose to have zero, one, or two blocking sides. For example, if you set the Z-axis to "+Z" then top of the grid will act as a ceiling. If you set it back to none, the top of the grid will act as open space, letting smoke exit freely.
GPU
Use CUDA Acceleration – FumeFX uses NVIDIA® CUDA to accelerate simulation calculations on the GPU. While GPU and CPU solvers produce different results, both rely on the same core simulation principles.
Performance improvements depend on grid size and simulation complexity, but testing has shown an average speed gain of approximately 2× to 5× over CPU-only simulations (NVIDIA RTX 4090 24 GB + AMD Ryzen 9 7900X test system).
For optimal results, we recommend an NVIDIA RTX 4090. The minimum suggested GPU for practical use is an RTX 3090.
Current GPU limitations:
– Sharpening
– Effectors
– NSim
- Spline Follow
– Burning with Oxygen
– Conservative Advection
– Wavelet Turbulence
– Fluid Mapping
Max GPU Memory % – At the start of the simulation, FumeFX GPU allocates the memory it will use for the entire run. If all three axes have Boundless set to None, FumeFX allocates only the amount needed for the full grid. If any Boundless option is enabled, FumeFX instead allocates a user-defined percentage of the available GPU memory.
Max CPU Memory % – FumeFX CPU follows the same logic. It allocates its memory once at the beginning of the simulation. When all axes have Boundless set to None, only the exact amount required for the full grid is allocated. If any Boundless setting is enabled, FumeFX reserves the user-specified percentage of available CPU memory.
During CUDA-accelerated simulations, the CPU memory limit should always be higher than the GPU memory allocation (wrong setup shown in image above). Otherwise, the GPU will be forced to reduce its memory usage, and you won’t be able to take full advantage of your graphics card.
Note: If a GPU simulation runs out of memory, FumeFX will seamlessly fall back to CPU simulation. This transition can be noticed in several ways: the GPU memory usage progress bar will turn dark red, the simulation output window will report the change, and the system taskbar will show increased CPU usage. To monitor GPU activity, select CUDA from one of the GPU graphs in the performance monitoring dropdown.
When simulating large grids near your GPU memory limits, we recommend turning off the GPU Viewport Display or setting its quality to 3. This helps reduce GPU memory usage and prevents unnecessary memory pressure during simulation.
Adaptive Grid
Adaptive Grid - While enabled, the grid will expand and shrink with the movement or absence of fluid to use only the necessary portion of the overall grid space. It is highly recommended to leave this enabled because it reduces simulation time and file size.
Allow Shrinking - Most of the time you will want grid to shrink and to occupy minimum required space for the simulation. However, if grid expands and shrinks during the simulation it will also delete velocity field which is very important to achieve natural rolling motion.
Extra Padding - Allowing velocity field to develop throughout the grid volume is of utmost importance for creating naturally moving fluid. When adaptive grid sensitivity is set to fields (smoke, fire...) adaptive grid will be created very tightly around smoke or any other field. Use this option to add extra voxels around the fields so that velocities can develop more naturally.
Sensitivity - FumeFX allows you to choose between different modes that will allow it to adjust grid size based on values near to the adaptive grid boundary.
Include Pressure - Enable this option if you want grid to expand based on pressure from the burning process and additional pressure introduced from sources.
Field Threshold - This parameter defines boundary sensitivity to smoke, fuel, fire and temperature. Larger number will generate tighter adaptive grid. If there is a source that doesn`t emit any matter and just produces velocities, use sensitivity by Velocity + Fields to be able to enlarge the grid.
Velocity Threshold - When Adaptive option is enabled and Sensitivity is set to Velocity + Fields, this parameter will define boundary sensitivity to velocity magnitude. Larger number will generate tighter adaptive grid. A good starting value would be grid Spacing area divided by a factor of 10.
Boundless - This option allows you to select grid axes that will be allowed to expand beyond the boundaries set by the container shape. Keep in mind that grid expansion will be limited by the amount of available computer memory.