Volumes use a large amount of memory and it’s important to keep your device’s limitations in mind.
When simulating with 32-bit fields, Axiom will use 1GB of memory for every 13m voxels you allocate. With a 8GB GPU you will be able to simulate 91m voxels under ideal conditions.
<aside> 💡 By default Axiom reserves 1GB of memory on your device as to not accidentally fully saturate it and lock up your computer.
</aside>
Your computer will usually be using memory for tasks other than simulating so you will typically have less than the total amount of device memory available to you.
If you are using Metal or Cuda your GPU has access to your system memory as well as its own. The solver will try to use all the available memory on your GPU before starting to use your system memory.
<aside> 💡 Please note that when the solver uses shared memory it will be slower as it needs to access memory over PCIe.
</aside>
It is recommended to not have a live render and Axiom simulation active at the same time. Unless you have a very large amount of GPU memory, doing both at the same time may cause the two processes to compete for resources and may result in instability.
Rendering volumes in Houdini’s or any other DCC’s viewport can use a large amount of memory. If you want to maximize the amount of memory that Axiom can use you can try reducing the viewport volume resolution or unload heavy assets when simulating.