Infatoshi

Infatoshi / craftax.c

Public

Pure C / AVX-512 port of Craftax-Classic. 47.8M SPS on a Ryzen 9 9950X3D -- 3.2x an RTX Pro 6000 Blackwell on the same env.

11
1
100% credibility
Found Apr 19, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

A optimized CPU version of the Craftax game environment that simulates thousands of parallel worlds faster than GPU alternatives for AI training.

How It Works

1
🔍 Discover Fast Game Simulator

You find craftax.c, a tool that runs tons of crafting game worlds super quickly on everyday computers.

2
📥 Get It Ready

You grab the files and set everything up with a simple build step.

3
⚡ Test the Speed

You run a quick check and see it handling millions of game moments every second.

4
🚀 Beat the Big Guns

You're amazed as your regular processor flies past even powerful graphics cards in game speed.

5
🔗 Link to AI Training

You connect it easily to your smart agent learning tools using ready-made helpers.

6
🎮 Train Smarter Players

You watch your AI characters learn crafting, building, and surviving way faster than before.

🏆 Lightning-Fast Wins

Your simulations run at peak speed, letting you train top-notch game AIs in record time.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is craftax.c?

craftax.c delivers a pure C port of the Craftax-Classic RL environment, optimized with AVX-512 for modern AMD CPUs. It simulates thousands of parallel game worlds—complete with Perlin terrain, mobs, crafting, and survival mechanics—at up to 47.8M steps per second (SPS) on a Ryzen 9 9950X3D. Python bindings integrate seamlessly with PufferLib for training agents on craftax classic benchmarks, swapping GPU dependency for CPU throughput.

Why is it gaining traction?

It crushes GPU baselines, delivering 3.2x the SPS of an RTX Pro 6000 Blackwell in the same env, proving CPUs excel at branchy, cache-sensitive sim loops. Custom thread pools and pipelined world generation hide resets, sustaining peak perf across NE=32k batches. Devs chasing craftax-classic records get plug-and-play speedups without CUDA tweaks.

Who should use this?

RL engineers benchmarking craftax-classic on AVX-512 desktops like Zen 5 rigs. PufferLib users training tiny policies (45k params) who prioritize SPS over GPU scale. Sim devs porting latency-bound envs to pure C for CPU clusters.

Verdict

Grab it for craftax classic benchmark dominance if you've got a 9950X3D—benchmarks verify the 47.8M SPS claim. At 10 stars and 1.0% credibility score, it's raw prototype territory with solid docs and sanity checks; test locally before production.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.