--- title: "GPU Distributed Rendering" canonical: "https://documentation.chaos.com/space/VMAX/113575281/GPU%20Distributed%20Rendering" format: markdown --- This page provides information about the Distributed Rendering rollout available under the Perf. tab when the production render is V-Ray GPU.    Overview This rollout in the Perf. tab deals with Distributed rendering. Distributed rendering is the process of computing a single image over several different machines. Before being able to use the distributed rendering option, you must determine the machines that will take part in the computations. V-Ray should be properly installed on those machines  and V-Ray GPU Render Server application must be running: either  as a service  or as a standalone application . For additional information on configuring and running distributed rendering, please refer to the  Set Up Distributed Rendering . UI Path: ||Render Setup window|| > Performance tab > Distributed Rendering rollout ( Renderer set to V-Ray GPU ) Parameters On  – Enables distributed rendering. Render servers  button – Opens the V-Ray Interactive renderer DR settings window. Add server  – Click to manually add a server by entering its IP address or network name. Edit server  – Click to change data about a render server. Remove server  – Deletes the currently selected server(s) from the list. Resolve  – Resolves the IP addresses of all servers. Use local host  – Determines whether V-Ray runs on the local machine when DR is used. When enabled, V-Ray runs on the local machine as well as on all machines included in the distributed rendering. When disabled, V-Ray does not run on the local machine and instead it only uses the render servers. Use this option when the local machine is needed for other computational tasks. Render region warning/test resolution warning 1  – When enabled, V-Ray displays a warning if the render region/test resolution options are enabled before sending a job to the network.  By default, the render server machine uses all its graphic cards in CUDA DR rendering, if nothing has been specified beforehand. Set up CUDA Devices for DR V-Ray GPU can run on one or more CUDA/RTX-capable graphics cards. To specify which cards are to be used, go to Windows Start Menu > Programs > V-Ray 6 for 3ds Max #### > Select rendering devices for V-Ray GPU for 3ds Max #### for each render server machine and pick. By default, the render server machine uses all its graphic cards in CUDA DR rendering, if nothing has been specified beforehand. The tool can also select GPU devices for denoising. It also allows for setting Low GPU Thread Priority per device. Note that when multiple devices are selected for denoising, out of all the devices that match the query only the device with the highest compute capability is used for the denoising process. The Select devices for V-Ray rendering tool has less functionalities and a different design in earlier versions of V-Ray. It is located again in Windows Start Menu > Programs > V-Ray for 3ds Max > Select rendering devices for V-Ray GPU for 3ds Max #### . Notes The VFB test resolution is ignored for command-line rendering and Backburner jobs. Test Resolution button is inactive in IPR if  Fit resolution to VFB  is enabled. Second Generation Distributed Rendering In future versions, V-Ray will install with the second generation of Distributed Rendering (DR2), which improves on the first generation with better scaling when adding many Render Server machines, better repeated render startup with new caching, and better utilization of high-bandwidth networks. See more details on the Distributed Rendering page. Some important details about using DR2 include: The default TCP/IP port numbers used in DR2 are different, compared to the "classic" (DR1) mode: V-Ray uses port 20208 by default. V-Ray GPU uses port 20209 by default. DR2 introduces the concept of a "dispatcher" node. The dispatcher handles the communication between the Render  Client and the Render Servers, allowing many more machines to effectively participate in the rendering. The Render Client can act as a dispatcher, but optionally any other machine can be dedicated to this role in the network. You can select a dedicated network dispatcher in the Distributed Rendering settings. In the event that a remote dispatcher is selected, you can list the servers with the additional -renderhost= argument on the dispatcher's side. More than one -renderhost can be specified. That way there is no need to explicitly add the servers on the client side, as the dispatcher will take care of the communication afterwards: