Lab tuning protocol, not a performance walkthrough. DCENT_OS has accepted-share evidence on exact S9 XIL hardware, but the current artifact still lacks a complete witnessed install, soak and recovery capstone. Do not use this page as authorization to flash production hardware.
What tuning evidence must answer
- Which exact S9 chassis, control board and hashboard revisions were tested?
- Which artifact hash, signature, Git commit and configuration produced the run?
- Were hashrate, reject rate, watts, J/TH, chip temperatures and fan RPM measured independently?
- Did the unit remain stable for 24 and 72 hours, including pool reconnect and thermal transients?
- Could the operator return to the recorded stock state using the documented recovery path?
Safe evaluation order
- Identify the exact Zynq/XIL board and exclude S9i/S9j assumptions.
- Rehearse recovery on sacrificial hardware before any DCENT_OS write.
- Capture a stock baseline under controlled inlet temperature and power conditions.
- Change one bounded parameter at a time. Never infer safe voltage, frequency or fan limits from another chassis.
- Stop on hardware errors, missing chips, unstable pool shares, thermal excursions or recovery uncertainty.
- Publish raw evidence and negative results; do not report a modeled target as measured performance.
No benchmark claim yet
D-Central is not publishing a DCENT_OS overclock, undervolt, hashrate or efficiency promise here. The page will gain measured results only when a named artifact and exact hardware lane have a reproducible evidence receipt.
Download the machine-readable benchmark template or use the reviewer kit to submit an independent reproduction or failure report.
When your miner retires: one more layer decentralized
Open-source firmware extends a miner's useful life. When a model retires from profitable mining, the power, cooling, and connectivity you built around it can still serve the distributed-compute layer — heat reuse, mesh comms, or freeing infrastructure for sovereign AI.
