Chip Binning

Chip binning is the manufacturing practice of testing each ASIC die and sorting it into performance "bins" — grades defined by attributes like maximum stable clock speed, the voltage needed to reach it, current leakage, and tolerance to temperature. No two dies emerge from a silicon wafer perfectly identical: microscopic variation in transistor threshold voltage, gate-oxide thickness, and interconnect resistance makes every chip electrically unique, even between neighbouring positions on the same wafer. This is the so-called "silicon lottery," and binning is how manufacturers turn that lottery into an orderly product line.

Why binning exists

Binning lets a foundry and its customers extract maximum value from each wafer. The fastest, lowest-leakage dies become premium parts; capable but less efficient dies are sold into lower-spec products at lower prices; and only genuinely defective dies are discarded. Rather than scrapping everything that misses the top grade, the manufacturer recovers revenue across a spread of performance tiers. Mining ASICs make the practice unusually visible: the same base design ships under multiple ordering codes — repair-bench chip testers, for example, distinguish variants like BM1366BS, BM1366BP, and BM1366AH, or the BM1370's AA/PA/PB families — and replacement chips must match the variant already populated on the board. On the repair bench, binning is not trivia; it is a parts-compatibility rule.

What it means for ASIC miners

For Bitcoin mining, binning quality directly shapes a machine's efficiency in joules per terahash and its real-world tuning headroom. Better-binned chips reach the same hashrate at lower voltage, running cooler and drawing less power — the difference between a unit that tunes happily and one that fights thermal throttling at the edge of its envelope. Because chips are supplied in voltage domains — groups sharing one regulated rail rather than each chip getting its own supply (see hash domain) — a single weak chip constrains its whole domain: the shared voltage must satisfy the worst die on the rail. This is also why two miners of the same model, same batch, can show measurably different wall power at identical hashrate, and it feeds the natural spread in the gap between nominal and actual hashrate.

Binning and autotuning firmware

Modern per-chip autotuning firmware is best understood as binning's second act, performed in the field. Instead of treating a board as uniform, the autotuner characterizes each chip's actual capability at runtime — the values are calculated on the machine, not preset at the factory — and assigns frequency accordingly, letting strong silicon work harder while weak silicon is driven gently. In effect, the firmware re-bins your specific dies under your specific cooling and settles on an operating point the factory's coarse grading could never find. The better the underlying silicon, the more an autotuner has to work with; the worse it is, the more an autotuner rescues.

Practical takeaways

When comparing machines, remember that model-level spec sheets describe bins, not your unit: expect a few percent of natural spread in efficiency between identical models, especially on the used market where the best-binned units may have been cherry-picked. When evaluating refurbished hardware or planning aggressive underclocking for quiet home operation, test the individual machine rather than trusting the label — its silicon, not its sticker, decides how far it tunes.

Browse chip specs in the ASIC chip database.