High-performance BIP39 mnemonic seed generator with GPU acceleration.
Generates BIP39 mnemonic seeds and searches for those with minimal character counts (spaces removed). Uses GPU acceleration to achieve billions of iterations per hour.
Key Features:
- GPU acceleration via NVIDIA CUDA
- Automatic resource throttling (80% CPU/GPU safety limit)
- Searches for seeds with < 46 characters
- No limit for rare seeds ≤ 42 characters
- Limit of 5 for seeds with 43-45 characters
- Real-time progress monitoring
cd rust
run.batcd python
run.bat- Speed: 1.5-2 billion iterations/hour
- Technology: Direct CUDA kernel integration with cuRAND
- Location:
rust/ - Setup: See
rust/README.md
- Speed: 1-1.5 billion iterations/hour
- Technology: PyTorch GPU acceleration
- Location:
python/ - Setup: See
python/README.md
Hardware:
- NVIDIA GPU with CUDA support
Software:
- CUDA Toolkit 11.8 or 12.x
- For Rust: Visual Studio Build Tools + Rust toolchain
- For Python: Python 3.8+ + PyTorch with CUDA
Results saved to:
mnemonics_log.txt- Detailed log with timestamps and iterationsseeds_output.txt- Simple list format (mnemonic + char count)
- 45, 44, 43 characters: Maximum 5 seeds per character count
- 42 characters or less: Unlimited (collects all unique seeds found)
Real-world performance with NVIDIA GPU:
- Rust: ~1,500,000 iterations/second
- Python: ~1,000,000 iterations/second
- 2 hours runtime: ~3 billion seeds tested
- Rust Setup & Usage:
rust/README.md - Python Setup & Usage:
python/README.md
- Use Rust if you want maximum speed and don't mind compilation
- Use Python if you want easier setup or need to modify code frequently
Both versions produce identical results and share configuration.
Educational and research purposes only.