The Nonce Space Crisis
Modern ASICs are running out of header entropy
~1s
32-BIT NONCE EXHAUSTED
16
BIP-320 NONCE BITS
7
nTIME BITS STOLEN
HEADER-ONLY MINING
In header-only mining, ASICs hash only the 80-byte block header without constructing transactions or recomputing the Merkle tree. A controller prepares the coinbase, Merkle root, and target — the ASIC just rolls header fields (
nNonce, nVersion, nTime). This minimizes logic and bandwidth inside the chip.THE ENTROPY PROBLEM
Modern ASICs exhaust the 32-bit
nNonce in roughly one second. BIP 320 added 16 bits of nVersion as extra nonce space, but it's not enough. Some devices have started stealing 7 bits from nTime for additional nonce space (Stratum V2 discussion). This creates conflicts with proposals like BIP-54.THE BIP-54 CONFLICT
BIP-54 (the Great Consensus Cleanup, by Antoine Poinsot building on Matt Corallo's earlier proposal) wants to fix the timewarp attack by forcing the coinbase transaction's
nLockTime to the current block height minus 1. Miners resist this because they use nLockTime as convenient extra nonce space. This BIP directly addresses that conflict.Matt Corallo's 24-bit Solution
Expand nVersion nonce space from 16 to 24 bits
THE BIP
Matt Corallo proposes reserving 24 bits of the
nVersion field as extra nonce space for miners, replacing BIP 320's 16 bits. Published on the bitcoin-dev mailing list.48-bit
CURRENT ENTROPY
56-bit
PROPOSED ENTROPY
256×
MORE ENTROPY
WHAT THIS MEANS IN PRACTICE
With 48 bits of header entropy, a modern ASIC exhausts the space in about ~1.5 seconds, requiring a new job from the pool controller. With 56 bits, that extends to roughly ~2.5 minutes — drastically reducing the coordination overhead between ASICs and controllers.
BIP 320 (CURRENT)
16 nonce bits in
nVersion. 48 total header bits. ~1.5s grind time. Miners still need nTime or coinbase nonce rolling.NEW PROPOSAL
24 nonce bits in
nVersion. 56 total header bits. ~2.5min grind time. Eliminates need for nTime abuse. Unblocks BIP-54.nVersion Bit Layout
32 bits: 3 fixed + 24 nonce + 5 signaling
nVERSION FIELD (32 BITS) — PROPOSED LAYOUT
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
Fixed (3 bits)
Nonce space (24 bits)
Signaling (5 bits)
SIGNALING STILL WORKS
The 5 remaining bits still allow 5 concurrent soft fork signals via BIP-8/BIP-9. In practice, Bitcoin has never had more than 2 concurrent signals. The current 29 signaling bits are massively over-provisioned. 5 bits is more than sufficient.
BACKWARDS COMPATIBLE
Miners can start using the nonce bits immediately without a soft fork or any consensus change. Non-mining nodes don't validate
nVersion beyond the basic version number check. BIP 320 already established the precedent of miners freely rolling version bits. This proposal simply expands the range.Impact & Context
UNBLOCKS BIP-54 (GREAT CONSENSUS CLEANUP)
The primary resistance to BIP-54's
nLockTime fix was: "don't consume coinbase nLockTime as a convenient extra nonce". With 24-bit nVersion nonce space providing ~2.5 minutes of grind time, miners no longer need coinbase nLockTime for nonce rolling. This removes the main objection to BIP-54.STRATUM V2 COMPATIBLE
The Stratum V2 mining protocol already supports nVersion rolling. The expanded range integrates naturally with existing mining infrastructure.
NO CONSENSUS CHANGE
Miners can adopt this unilaterally. No soft fork, no node upgrades, no activation mechanism. Just start rolling more bits. This makes deployment trivial.
MATT CORALLO
Long-standing Bitcoin Core contributor, key contributor at Spiral (formerly Square Crypto), creator of the Lightning Development Kit (LDK), and developer of the original FIBRE block relay network. When Corallo proposes a mining BIP, the mining industry listens.
WHY NOW?
BIP-54 is now very mature and close to activation, but has two outstanding criticisms: (1) restricting 64-byte transactions is ugly, and (2) the coinbase
nLockTime fix reduces available nonce space. This BIP removes criticism #2, clearing the path for the Great Consensus Cleanup to proceed.