Mining rig maintenance: dust, thermal paste and a simple schedule
A rig does not need daily attention — that is its charm. But "set it up and forget it for two years" always ends the same way: temperatures creeping up for no visible reason. The reason is always the same — dust and aging thermal interfaces. The good news: all rig maintenance fits on a one-page schedule.
The short version (for the no-time-to-read)
- Dust is a blanket on the heatsink. A year without cleaning easily adds 5–15 °C at the same wattage.
- Blow out every 1–3 months (depending on how dusty the room is) — cheaper than replacing fans and pads.
- Thermal paste lives 2–3 years, memory pads about the same. A card older than three years with no service is almost certainly running hotter than it should.
- Hold the fan blades while blowing. An air stream overspins the fan: bad for the bearing, and the motor acts as a generator feeding voltage back into the card.
- A temperature log is the best alarm. An upward trend at unchanged settings means: grab the air can, not a new card.
What dust actually does
Dust harms three ways. First, a layer on the heatsink fins acts as insulation: heat leaves worse, temperatures rise at the same watts. Second, dust works into fan bearings and accelerates wear — and fans are already consumable number one (how a fan predicts failure). Third, dust plus humidity makes a conductive film on the board; rare, but spectacular.
The indirect effect is worse than the direct one: a hotter card → higher fan speeds → a louder rig and faster wear → even worse cooling. Regular cleaning breaks this loop at its cheapest point.
A one-page maintenance schedule
- Every 1–3 months — blow-out. The interval depends on the room: a balcony with a filtered intake — quarterly, a garage — monthly. A simple marker: card temperatures creeping up at unchanged settings — time to clean.
- Monthly — a glance at RPM and connectors. Fan speeds rising at familiar temperatures means a tired fan. Check power connectors for discoloration while you are at it.
- Every 2–3 years — thermal paste and pads. Paste dries and cracks, pads lose elasticity. The symptom is a growing gap between core and memory temperatures (on hot GDDR6X — here).
- Yearly — a general inspection. Snug the connectors, look over the cables, make sure the risers sit tight (on their quirks — a dedicated article).
How to clean it right
- Rig off and unplugged. Blowing "hot" turns dust into a conductive fog inside running hardware.
- Compressed air, not a vacuum. A vacuum cleaner on a live board is a static generator. Use a can or a compressor with short bursts.
- Hold the fans with a finger or a zip tie. The air stream overspins the rotor past its rated RPM: the bearing suffers, and the motor works as a generator pushing voltage back into the board.
- Outside, not around the room. Blowing a rig out in a living room just relocates dust from the heatsinks to your lungs. Balcony, street, garage.
- Measure after cleaning. Before/after temperatures at the same settings are the best proof the work was worth it. Usually minus 5–10 °C.
Where automation fits
A temperature trend is invisible day to day — it shows in history. OCBot watches the rig around the clock: every card's temperatures and fan speeds are recorded continuously, so a slow temperature rise at unchanged settings is visible in numbers long before it becomes audible. Cleaning the rig on an alarm is nicer than replacing fans after a failure.
This article is general information summarizing public practice; actual intervals depend on your room and hardware. OCBot is an independent project: not affiliated with NVIDIA, HiveOS or Hiveon. Use at your own risk — see the disclaimer.