Advanced Job Paramaters: Minimum VRAM & Maximum GPUs

Guide to using Advanced Job Settings on the Render Network

The Render Network Job Configuration page offers more granular controls to help you more optimally leverage the Render Network diverse pool of GPUs. Although they are optional, these Advanced Job Parameters are recommended for many scenes in order to increase the efficiency of your renders.

Setting Minimum Node VRAM

If your scene is resource-intensive, it is recommended to select the appropriate VRAM requirement in order to ensure it stays within GPU VRAM limits. This will reduce failures, job re-queuing, and increase performance on complex scenes that require high amounts of VRAM. The setting is is particularly relevant for Octane users, where performance is optimized for VRAM and not Out-of-Core Geometry.

By effectively setting Minimum Node VRAM on complex scenes, you can prevent your render from utilizing Out-of-Core Geometry, which will slow down Render times and increase failure rates, and can lead to higher costs because Render Network nodes are not configured for Out-of-Core Geometry.

Figure 1: Setting Minimum Node VRAM

When setting Minimum Node VRAM, you can select between 6, 8, 11, & 16GB Minimum VRAM Settings. If Minimum VRAM is unspecified, the assigner will select nodes algorithmically. It is recommended to include GPU headroom for your scene. For example, if your scene is 7.4gb VRAM, it is recommended to select 12GB VRAM or higher rather than selecting 8GB VRAM, which does not have a lot of headroom. You can check your memory usage in through the Statistics Panel or the Rendering Window.

Setting Maximum Number of GPUs

On the Render Network, you can define the Maximum Number of GPUs per node you would like to render on a scene. For example, if you set the Maximum GPU setting to 2 GPUs, your job will only be rendered with nodes with up to 2 GPUs. Using this feature on shorter animations or less complex scenes, will lower job costs, restricting a job from being sent to large multi-GPU setups that can charge more.

Figure 2: Setting Number of GPUs

For jobs using Redshift or Blender Cycles, the default Maximum GPU setting is 4 GPUs, but you can choose anywhere between 1-4 manually. For jobs being rendered in Octane, the default is unlimited, but in practice, this means on Tier 2 the jobs will go to nodes with up to 14 GPUs. If you know that your Octane scene is light and fast, it is better to restrict the number of maximum GPUs, which can be set between 1-8 GPUs, allowing you to save on Octane Bench usage.

Figure 3: A breakdown of Maximum GPU settings in different 3D tools

The addition of a Maximum GPU setting on Render allows artists to customize their rendering preferences to meet the specific demands of a job, enabling them to take advantage of the wide variety of Render Network node configurations and granularly optimize between cost, performance, and speed.

When artists optimally set their job settings to efficiently use VRAM and Maximum Node GPUs, it also helps Node Operators, because it enables the network to more effectively allocate high performance jobs to powerful nodes, while also providing fast and cost effective rendering for less complex scenes. Additions to Minimum VRAM and Maximum GPU settings may be updated over time, as GPU hardware changes and artist usage evolves.