We are in the process of building out our Horde infra inclusive of running HLOD generation through Horde.
Looking to get info on what instance types in AWS Epic currently uses for HLOD and also in general what the requirements are for HLOD generation on Horde agent systems.
Hello!
Internally, the requirements for our HLOD builds varies a bit and depends on each project needs.
Fortnite HLODs are built daily, using 5 instances of r6a.8xlarge
City Sample is built on demand only, as its content doesn’t change anymore. It uses 10 g4dn.8xlarge instances.
You can probably get away with AWS instances that have no dedicated GPU, by providing the -AllowSoftwareRendering argument. In some cases however you may run into driver issues that will either cause the build to crash or will give incorrect results…
Tweaking your HLOD layers settings can help lower the memory requirement in case that’s an issue for you. City Sample ships with 3 different HLOD layer setup, and at the time, using the lowest quality allowed a build to be performed on a machine with 64GB of ram.
Here are some more resources on HLOD & build distribution using Horde.
Let me know if you have more questions!
Regards,
Sebastien