Replication Graph For Optimizing Real-Time Strategy Games
summary
Shows concrete wins: using Unreal’s Network Profiler with a deterministic 30s capture (20 units per player, skip first 2s) finds trees costing ~1,600 KB. Changing gameplay tags from full strings to index replication drops tree traffic to ~500 KB and tag bytes from ~1,300 to ~60, toggling BReplicateMovement cuts another chunk. Migrating to Replication Graph (the tech Epic built for 50,000 Actors to 100 connections) and a vision-based node that replicates only seen units yields large bandwidth and CPU reductions; code and nodes are on GitHub.
AI-written summary · about
chapters (13)
- 0:00RTS game optimization intro
- 2:36Test scenario setup
- 4:23Network perf in Unreal Engine
- 9:51Bandwidth bottlenecks in trees
- 11:14Gameplay tag replication
- 12:43Static object movement off
- 14:55Excluding init spikes
- 16:22Replication Graph scaling
- 19:38Relevancy checks improved
- 22:31Basic Replication Graph setup
- 24:41Production readiness
- 26:53Vision-based nodes
- 31:21Open-sourcing Replication Graph
related talks
43mOptimizing animations in Unreal Engine 4
52mOptimising your game AFTER it ships
01h 00mUnreal Engine Game Optimization on a Budget
51mOptimising Games for a Complex World
37mSpeeding up Game Logic in Unreal Engine
01h 03mMaximizing Your Game's Performance in Unreal Engine
36mImproving Developer Iteration in Unreal Engine 5
50mImproving Iteration in Geo-Distributed Scenarios