Ulsan · South Korea · Since 2018
When a miner fails to boot, every hour of downtime costs hash rate
We assess ASIC miners at 0.05 TH/s and above — measuring cold-start reliability, thermal recovery after power events, and sustained uptime across boot cycles.
Request an uptime assessmentWhat we measure on every rig
Cold starts, warm restarts, and recovery after brief power dips — logged across 72-hour observation windows.
Hash rate stability, fan response, and thermal throttling during continuous operation at rated wattage.
We focus on hardware that meets or exceeds 0.05 TH/s nominal output — the threshold where boot failures become costly at scale.
Assessments for operators who cannot afford silent failures
Small hosting rooms in Nam-gu and larger farms along the Ulsan coast rely on miners that restart cleanly after maintenance windows and grid fluctuations. Our reports document which models recover without manual intervention and which require repeated power cycles.
Each assessment includes boot-cycle logs, thermal readings at intake and exhaust, and a plain-language verdict on whether the unit is suitable for unattended operation.
How we work
Assessment types
Boot Cycle Diagnostics
Focused 24-hour protocol isolating restart failures — ideal when a unit hashes but refuses clean cold boots.
Fleet Comparison Report
Side-by-side boot reliability and uptime scores for up to eight miners in a single documented comparison.
Uptime & Boot Reliability Assessment
Flagship 72-hour evaluation of cold starts, warm restarts, and sustained hash rate for miners at 0.05 TH/s and above.
From recent field notes
Why the 0.05 TH/s Floor Matters for Boot Testing
Below certain hash rate tiers, boot failures are annoying. Above them — especially near 0.05 TH/s and higher — each failed restart carries real revenue loss at scale.
Coastal Humidity and Miner Uptime in Ulsan
Intake moisture correlates with boot failures we log at coastal hosting sites — patterns from summer assessments along the Ulsan shoreline.
"The boot-cycle log for our S19 batch showed three units that needed two power cycles before hashing. We relocated those to a monitored row — the report named the serial ranges, not vague warnings."