Articles on mining hardware

Operation and reliability

• Why a mining rig freezes solid — and what to do about it

  Power delivery, risers, hung processes, RAM-only logs. A diagnosis checklist for a silent rig and what is worth handing to automation.

• GPU temperature and lifespan: what the physics actually says

  Black's equation (electromigration), Coffin–Manson (thermal cycling), the 10 °C rule. Why "letting the card rest" does nothing and what power limits have to do with it.

• GPU fan as a failure predictor: what noise and RPM tell you

  The fan dies before the card — and warns in advance: creeping RPM at the same temperature, bearing rattle. How to catch it by charts and by ear.

• Mining rig risers: why they are the usual suspects in "random" failures

  Riser power (SATA vs 6-pin), cheap USB cables, seating in slots: how a lunch-priced part turns overnight drop-offs into a pattern — and how to rule it out.

• GDDR6X memory: why it runs hotter than the die and what to do about it

  Sensors read 70 while memory works at its ceiling. The GDDR6X hotspot, thermal pads, and how memory clocks map to heat.

• Should you switch a mining farm off at night? Physics and money

  Every shutdown is a thermal cycle and a cold start. Night savings versus night revenue: when stopping is justified — and when it is the most expensive "saving".

• Mining rig maintenance: dust, thermal paste and a simple schedule

  Dust is a blanket on the heatsink: +5–15 °C after a year without cleaning. Blow-out every 1–3 months, paste and pads every 2–3 years, a temperature log as the alarm clock.

• A mining rig in the heat: how to survive summer without losses

  Every +10 °C in the room is +10 °C on the cards. Exhaust beats air conditioning, a summer mode with reduced watts, and the fall inspection.

• Home mining farm noise: where it comes from and how to cut it

  Noise = fans = watts. Why minus 100 W per card beats a quiet cooler, bearing rattle as a symptom, soundproofed boxes with exhaust and rack resonance.

• Wi-Fi or Ethernet for a mining rig: what actually matters

  Mining needs kilobytes, not gigabits: latency and packet loss decide. Ethernet vs Wi-Fi vs USB dongles, stale shares and a backup link.

• Relocating a mining rig: balcony, garage, or a rented space

  Humidity, dust and temperature at the new site; condensation on cold hardware, wiring and network before the move, and safe card transport.

• Throttling: why a GPU lowers its own clocks

  Thermal and power throttling: how to tell them apart on charts, what fixes each, and why a voluntary minus 10% of watts beats a forced one.

• Rig sees 5 of 6 GPUs: a checklist for finding the missing one

  A missing card is almost never a dead card. Power, riser, slot, driver and the swap test as the verdict: find the culprit in one evening.

• Updating miners and drivers: when to do it and when to leave it alone

  Update for a reason, not on a schedule; the one-rig rule, a backup and a rollback path, official sources only.

Power and budget

• A PSU for a 6–8 GPU farm: why "just enough" is the most expensive saving

  The 80 PLUS efficiency curve, transient load spikes and ATX 3.x, rails and OCP, one cable per card. Watt headroom is not money lying idle.

• Home wiring for a mining rig: outlets, cables and breakers

  A 2 kW rig is a permanent 24/7 load. One outlet per rig, why extension cords overheat, a dedicated circuit with its own breaker, a warm plug as an alarm.

• A UPS for a mining rig: do you need one, which type, and when it pays off

  A UPS guards against one-second flickers, not long blackouts. UPS types, watts vs volt-amps, pure sine wave, and when a voltage relay is enough.

• Your mining rig's power bill: where the number comes from and how to verify it

  Wall draw exceeds the sum of GPU watts: PSU efficiency, overhead, seasonality. A smart plug, time-of-use metering and an honest number for the profitability math.

Buying hardware

• Buying a used GPU for mining: how to avoid a dead one

  The "killed by mining" myth, an inspection and 30-minute test checklist, brands and revisions, and why Samsung/Micron/Hynix memory beats the logo on the shroud.

• Breaking in a new GPU: the first two weeks in your rig

  The bathtub failure curve, stress tests at stock clocks, seals untouched until burn-in ends, and screenshots as insurance: let hidden defects surface while returns still work.

Tuning and money

• Manual vs automatic overclocking: the arithmetic of labor

  Tuning hours per card, stability runs, repeated work after every algorithm and driver switch. A comparison by arithmetic, not by advertising.

• Mining pools: fees, payouts, and what is subtracted between rig and wallet

  Pool fees, PPS vs PPLNS, min-payout thresholds and network fees: how much actually reaches the wallet and how to compare pools by facts, not by ads.

• The mining winter: what to do when profit turns negative

  The price drops and per-card revenue sinks below its electricity. Honest arithmetic, efficiency mode instead of chasing hashrate, when the rig should sleep — and why not to sell at the bottom.

• How to compute mining profitability honestly

  Revenue is gross; profit comes after electricity and fees. Formulas, a worked example, the classic self-deceptions and the calculator with your own tariff.

• Pool hashrate lower than rig hashrate: where the percent goes

  Reported vs effective: averaging windows, pool luck, stales and rejects. Which losses are normal, which are a red flag, and how to compare by days.

• Network difficulty: why revenue drops at the same hashrate

  Your slice of the pie is your hashrate divided by the network's. The network chases the price upward and coin revenue melts by itself. How to track it.

• Buy the coin or build a rig: an honest comparison

  The coin is a pure price bet; the rig is the same bet plus cash flow and resale value, minus expenses. Scenario arithmetic and break-even points.

• A second mining rig: what actually changes when you scale

  What doubles fully (heat, noise, wiring), what scales for free (knowledge, presets, spares), and the checklist before rig number two.

• Core or memory: what to tune for different mining algorithms

  Memory-bound vs core-bound algorithms: which knob drives hashrate, why no universal preset exists, and what watts spent in the wrong place cost.

• Testing overclock stability: why ten minutes is not a test

  A short test only catches gross overshoot. A 24-hour logged run, the signs of instability, and when a preset must be re-verified.

Where to next

A new article goes live about once a week. If you want to suggest a topic — drop us a line; contacts are on the homepage. The tool these articles grew out of is OCBot: per-card auto-overclocking, the rig's power budget and 24/7 monitoring.

• Card and preset catalog
• Profitability calculator
• Try OCBot