AI summary by TwinMind
Making Minecraft 100x faster (by rewriting it in Rust)
Theo - t3․gg
17m 1s
Personal Background & Introduction
- Speaker's history with Minecraft development
- Started hosting servers in high school
- Led to career as developer at Twitch
- Focus on Java Minecraft server performance issues
- Bedrock C++ rewrite critique: riddled with bugs, weird behaviors, not better performance
- Java server hosting performance has gotten worse over time
Pumpkin Server Performance Claims
- New Rust-based Minecraft server project
- Dramatic performance improvements vs Java servers:
- Java baseline: 2GB RAM, 24% CPU usage (10 players)
- Pumpkin: 27MB RAM, 1.5% CPU usage (10 players)
- ~100x reduction in memory usage
- 20x reduction in CPU usage
- Trade-off: written in Rust, not Java
Sponsor Segment - Blacksmith
- GitHub Actions alternative for faster, cheaper builds
- Tokyo build example:
- GitHub Actions (2v CPU): 6 minutes 12 seconds, ~6 cents
- Blacksmith 2v CPU equivalent: 3 minutes 51 seconds, <2 cents
- Blacksmith 4v CPU: 2 minutes 47 seconds, <2 cents
- Node.js compilation: half the time, 1/3 the cost
History of Rust Minecraft Servers
- Website tracking: "Days since last Rust Minecraft server" - 37 days since FMC release
- Timeline of projects:
- 2019: Feather (first)
- June: MCHPRS
- Graphite
- Veilance
- 2-year gap
- Hyperion (still maintained, commits hour ago)
- FMC (currently in rewrite)
Pumpkin Project Details
- Built entirely in Rust
- Goals: security, compatibility, performance
- Explicit non-goals:
- No plugin/mod compatibility with other servers (CraftBukkit, Fabric)
- Not intended as framework for building other servers
- All-in-one solution approach
Feature Status & Limitations
- Missing critical features:
- World borders
- World saving
- Light recalculation (everything appears dark)
- Supported features include block placing/breaking
- Developer response on lighting: "everything is dark, we're working on it"
- Target: enable low-end hardware hosting (Raspberry Pi testing planned)
Technical Challenges
- Minecraft server standard not open source
- 10+ years of reverse engineering Java implementation
- Protocol documentation shows complexity:
- Data types implementation
- Request types (set block, destroy stage, block updates, boss bars, etc.)
- Redstone logic particularly complex
- Command blocks require server-side execution
Live Testing Results
- Local testing with cargo run --release
- Resource usage observations:
- Idle: 4MB RAM, 0% CPU
- Connected player: 200MB RAM, 0.01% CPU
- Chunk generation spikes briefly, then drops
- Lighting issues visible (broken lighting system)
- Comparison to typical Java servers: 8GB+ RAM usage
Redstone Complexity Examples
- Single-player Minecraft runs as local server (changed in version 1.3)
- All Redstone logic runs server-side
- Extreme example: Pokemon Red/Blue recreated in Minecraft
- Entire Game Boy cartridge reimplemented
- Requires MCHPRS (Minecraft High Performance Redstone Server)
- 10,000x speed boost needed for playability
- Minecraft-in-Minecraft implementations exist
Performance Comparisons
- MCHPRS benchmarks: 9% CPU, 400MB RAM
- Pumpkin: 1.5% CPU, 27MB RAM
- Demonstrates significant advancement in Rust implementations
Current State & Future
- Minecraft server hosting in "Renaissance period"
- Multiple competing implementations
- Plugin ecosystem fragmentation across 15+ different platforms
- Performance gains vs compatibility trade-offs
