Unreal Engine cook pipeline optimisation

Principal Consultant · 2026 · 3 months · 2 min read

Reduced iteration time on asset cooks by profiling bottlenecks and restructuring the pipeline.

Overview

Contract engagement to improve Unreal Engine asset cook times for a games studio shipping on multiple platforms. The studio's content team was blocked by long cook cycles during iteration.

Problem

Full and incremental cooks were taking too long, slowing content iteration and blocking artists and designers from validating changes. The root causes were unclear — suspected shader compilation, texture processing, and redundant work across cook stages.

Constraints

  • NDA — client and title not disclosed
  • Could not change artist workflows or source asset formats
  • Had to work within the studio's existing Unreal version and build farm
  • Changes needed to be low-risk and reversible

Approach

Profiled the cook pipeline end-to-end with Unreal Insights and Superluminal, mapped time spent per stage, then targeted the highest-cost operations. Worked incrementally — each change measured before moving on.

Key Decisions

Profile with Unreal Insights and Superluminal before changing code

Reasoning:

Cook pipelines have many stages; guessing at bottlenecks wastes time. Instrumentation gave a ranked list of where minutes were actually spent.

Alternatives considered:
  • Add logging and timers manually
  • Rely on studio anecdote about slow stages

Optimise incremental cook paths before full cooks

Reasoning:

Day-to-day iteration uses incremental cooks. Fixing those gave immediate payoff for the content team.

Alternatives considered:
  • Focus on full cook time for CI builds first

Tech Stack

  • C++17
  • Unreal Engine
  • Unreal Insights
  • Superluminal

Result & Impact

Identified and removed several redundant processing steps. Incremental cook times dropped enough for the content team to iterate without leaving their desks. Delivered a profiling playbook the studio could reuse on future projects.

Learnings

  • Cook performance problems are rarely a single hotspot — they are usually a chain of small redundancies
  • Unreal Insights plus a sampling profiler gives enough signal to prioritise without weeks of instrumentation work
All projects