+34 641 10 27 16

Chip wear and stability under ASIC overclocking

Firmware documentation

Wear from overclocking: what the terahashes cost

No sales talk here. Overclocking is work beyond the factory mode, and it has a price. The point is to pay it knowingly.

What actually wears out

Silicon degrades from heat and voltage, and heat counts for more here. A chip that runs for years at 60 °C will outlast a chip running at 80 °C by a wide margin. Hence a simple conclusion: overclocking with good cooling is far safer than moderate overclocking in a stuffy room.

The second thing that suffers is the power supply. Running at the power limit shortens the life of the capacitors, and a degraded supply starts to sag on voltage, which brings errors on the chips. The circle closes.

Signs that an ASIC is running on the edge

  • The error rate is above 0.5% and slowly climbing.
  • The firmware drops frequencies on individual boards more often than it used to.
  • Occasional restarts in the logs with no clear cause.
  • Fan duty has gone up at the same room temperature.
  • Efficiency in J/TH creeps up month after month.

Any one of these signs means it is worth stepping down one profile. Losing a couple of percent of hashrate is cheaper than replacing a hash board.

Chart of rising errors and temperature on a maximum profile
Running on a maximum profile: errors rise before the hashrate falls

What matters for stability

FactorWhy it matters
Power supply headroomPast the limit the supply goes into protection, and the firmware sees chains drop out
CoolingTemperature directly sets how fast the chips degrade
Mains qualityVoltage sags give errors that look like a chip problem
Ambient conditionsDust, humidity and temperature swings matter more than they seem

On balance

In most cases a profile with a bit of headroom brings in more money than the maximum one. Count it this way: an ASIC on the maximum profile puts out 4% more, but once every two days it goes into a restart and loses 20 minutes. The arithmetic does not favour the record.

One more honest detail: autotuning is not all powerful. Some chips stay unstable, and you either live with a lowered frequency on them or turn the overall settings down. That is a normal outcome, not a sign of bad firmware.

Smooths out the swings Automatic profile switching by temperature and voltage adaptation in the advanced settings make life noticeably easier for an ASIC where conditions drift between day and night.

Quick answers

How much does overclocking shorten service life?
Nobody will give you an exact number, there are too many variables. Rule of thumb: temperature counts for more than frequency, at 65 °C an overclocked ASIC lasts a long time.
Can you overclock a used ASIC?
Carefully, and with an autotuning run. Worn chips overclock worse, and the error threshold comes sooner.
The firmware lowers the frequency by itself, is that a fault?
No, that is normal behaviour. The global frequency sets the target, the firmware holds stability.
Written by , ASIC repair and tuning engineer at asic.es.