--- title: "Sparse Caches and Instances of Simulators during Rendering" canonical: "https://documentation.chaos.com/space/PHX4MAX/124760432/Sparse%20Caches%20and%20Instances%20of%20Simulators%20during%20Rendering" format: markdown --- ## **Memory Usage** --- In Phoenix, if you load the same cache frame into several Simulators (via the [Input rollout](https://docs-chaos.atlassian.net/wiki/spaces/PHX4MAX/pages/124662003)), the cache's data is automatically shared between the Simulators and does not use extra memory. This holds for both copies and instances of the Simulator. If you render several Simulators in **Volumetric** or **Volumetric Geometry** [render mode](https://docs-chaos.atlassian.net/wiki/spaces/PHX4MAX/pages/124396558), there are a few options that would use extra memory per each Simulator, regardless if it's a copy or instance. Using these settings might also overflow your RAM if you are loading large sparse cache files: - Using [Volume Light Cache](https://docs-chaos.atlassian.net/wiki/spaces/PHX4MAX/pages/125312994) would use extra memory per each Simulator instance if rendering with V-Ray or Scanline. This extra memory is allocated as a dense grids, so remember to disable this option when rendering large sparse VDB caches. Please refer to the [Volumetric Rendering In-Depth](https://docs-chaos.atlassian.net/wiki/spaces/PHX4MAX/pages/124363172) for tips on how to render quickly even with **Volume Light Cache** disabled - most importantly, use V-Ray's **Progressive Sampler** and enable **Probabilistic Volumetrics**. - Using [Grid-based Self Shadowing](https://docs-chaos.atlassian.net/wiki/spaces/PHX4MAX/pages/124629510) would also use extra memory per each Simulator instance if rendering with V-Ray or Scanline. Again, the extra grids are dense, so change this option if you want to render large sparse VDB caches. You could switch to **Ray-traced Self Shadowing** instead, or if you are using GI, you could turn **Create Fire Lights** off entirely and rely on the GI to transport the illumination from the fire. - Using [Optimize Big Volumetric Grids](https://docs-chaos.atlassian.net/wiki/spaces/PHX4MAX/pages/125056277) would consume extra memory per each Simulator instance if rendering with V-Ray or Scanline. Again, the extra grids are dense, so disable this option if you want to render large sparse VDB caches. ## **Render Speed** --- Using [Grid-based Self Shadowing](https://docs.chaosgroup.com/display/PHX3MAYA4EDIT/Fire+Lights) or [Optimize Big Volumetric Grids](https://docs.chaosgroup.com/display/PHX3MAYA4EDIT/Smoke+Opacity) would also add a rendering pre-pass stage for each of the Simulators in the scene. All important tips for fast rendering of single volumes apply to many instances as well, please check the [Volumetric Rendering In-Depth](https://docs-chaos.atlassian.net/wiki/spaces/PHX4MAX/pages/124363172) for more information. ## **V-Ray GPU** --- Note that if you have many overlapping volumes, V-Ray GPU can render an infinite number of volumes in the scene, but no more than 4 volumes that overlap at the same position in space.