seine

Why seine

• Fast on whatever you own. Hand-tuned Argon2id kernels for AVX2, AVX-512, Arm NEON/SHA3 and NVIDIA Blackwell. Seine benchmarks them while it mines and keeps the fastest for your machine.
• Zero configuration. Run ./seine, paste your address, and you're mining in seconds. Threads, kernels, memory and CPU placement tune themselves.
• CPU and GPU together. Every core and every NVIDIA GPU mines at once.
• Pool or solo. Mines on bntpool.com out of the box, works with any Stratum pool, or solo against your own Blocknet node.
• Built for rigs and fleets. NUMA-aware placement, huge pages, a local control API with live events, and Prometheus metrics.

Windows setup guide · Connect to bntpool.com · Mining locally but pool hashrate is zero?

Get started

1. Download the archive for your platform from the latest release:

Platform	Archive	GPU
Windows x86_64	seine-windows-x86_64.zip	NVIDIA (drivers only, CUDA runtime bundled)
Linux x86_64	seine-linux-x86_64.tar.gz	NVIDIA (driver + NVRTC), AMD Radeon RDNA (ROCm HIP runtime)
Linux ARM64	seine-linux-aarch64.tar.gz	CPU only
macOS Apple Silicon	seine-macos-aarch64.tar.gz	CPU; Metal GPU for BLAKE2b blocks
macOS Intel	seine-macos-x86_64.tar.gz	CPU only
HiveOS rigs	seine-hiveos-<version>.tar.gz	NVIDIA (NVRTC bundled), AMD; see the HiveOS guide

The Linux builds run on glibc 2.31 or newer: Ubuntu 20.04, Debian 11 and later.

2. Run it:

./seine

3. Enter your Blocknet address (or @shortname) when prompted, and accept the default pool. That's it: seine starts hashing straight away and keeps tuning itself as it goes.

Each CPU thread needs about 2 GB of RAM. Seine sizes itself to your machine automatically. More options: setup and configuration.

Performance

97.1 H/s on the AMD EPYC 9355P with default settings, and 75.0 H/s on the EPYC 9554P with 64 mining threads and 1 GiB huge pages. Measurements and settings.

Every number is a measured result from repeated runs. Results for every CPU tested: performance results.

Get the most from your hardware

Seine is fast out of the box. These guides cover the last few percent:

Hardware	Guide
Linux / WSL	Huge pages: +4–8% on CPU
Windows	Large pages and CPU placement
EPYC, Xeon, many-core servers	NUMA, L3 groups and memory headroom
Apple Silicon	Threads, scheduling and Metal
NVIDIA GPUs	Requirements, multi-GPU and tuning

Documentation

• Setup and configuration: first run, flags, dashboard controls, shortnames, dev fee
• How automatic tuning works: thread counts, CPU kernels, memory recovery
• Solo mining: mine against your own Blocknet node
• HiveOS: add seine to a flight sheet with one installation URL
• Farm proxy: many rigs, one pool connection
• Control API: REST, live events and metrics for dashboards (OpenAPI)
• Flag reference: every command-line option
• Building from source
• Changelog
• For contributors: architecture and agent notes

License

Privacy policy.

BSD 3-Clause.