ASICBoost

ASICBoost is a proof-of-work optimization, described by Timo Hanke and Sergio Demian Lerner in 2016, that exploits a structural property of Bitcoin's double-SHA256 hashing. The 80-byte block header is hashed in two 64-byte chunks, and the expensive first-chunk computation — the midstate — depends only on the header's first 64 bytes. Arrange your work so that multiple candidate headers share those first 64 bytes, and one midstate computation serves many hashing attempts. Fewer total operations per hash attempt means less energy per hash — a direct improvement in joules per terahash, with published estimates of the saving in the double-digit-percent range depending on implementation.

Covert versus overt

There are two ways to obtain headers that share a midstate, and they differ enormously in their consequences. Covert ASICBoost keeps the version field untouched and instead grinds the transaction side: by manipulating transaction ordering and the tree structure, the miner hunts for Merkle root values that collide in the last 4 bytes of the header's first chunk. It is invisible on-chain — and toxic. The grinding creates perverse incentives, including a preference for smaller or emptier blocks, and the technique is fundamentally incompatible with the witness-commitment structure that SegWit places in the coinbase transaction — the collision grinding needs freedom over the transaction tree that the commitment removes. The suspicion that covert ASICBoost profits motivated some miners' opposition to SegWit was a live controversy of the 2017 scaling era. Overt ASICBoost takes the honest path: it rolls spare bits in the header's version field to generate header variety, is plainly visible on-chain, and carries none of the harmful incentives.

How overt boost is standardized

Overt ASICBoost rides on version rolling (BIP 310), the Stratum extension through which a miner and pool negotiate which version bits the miner may modify. The client requests a mask via mining.configure; the pool answers with the bits it permits; the effective mask is the intersection — conventionally within 0x1fffe000, sixteen bits that consensus rules treat as non-signaling. Those 16 rollable bits multiply the 32-bit nonce space by 65,536, giving hardware 2⁴⁸ attempts per job without touching the extranonce — and each rolled version corresponds to a distinct precomputed midstate inside the chip. BIP 310's standardization of overt rolling was the industry's direct answer to the covert controversy: same physics, none of the pathology.

Practical relevance today

Version rolling is now simply how modern mining works. Contemporary ASICs support overt boost as a matter of course, and it is enabled through the pool's version-rolling negotiation rather than any special hardware mode — both Stratum V1 (via the BIP 310 extension) and Stratum V2 (natively, through version-rolling fields in its job messages) carry it. Because the covert form clashes with SegWit-era block structure, overt rolling is the only practically relevant variant on Bitcoin today. If you run a miner, you are almost certainly using ASICBoost right now without having configured anything — a once-controversial patent-war battleground quietly absorbed into the protocol's plumbing, which is about the best ending an optimization can have. The patent history reinforced the lesson: ASICBoost was patented by its inventors, accusations of covert use distorted hardware politics for years, and the eventual placement of the technique under a defensive patent pledge helped clear the way for open, universal adoption. For miners today the checklist is short — use a pool that supports version rolling (effectively all major pools do) and confirm your firmware negotiates it, which any modern image does by default.