Max Registrations Per Block
Max registrations per block is a per-subnet hyperparameter, listed in the documentation as
MaxRegistrationsPerBlock, that once named a per-block ceiling on neuron registrations. The
reference now states it is not used for neuron registration; live admission is governed by
continuous burn pricing through burn half life and burn increase mult instead
(Subnet Hyperparameters).
What It Controlled
Under older registration models, the hyperparameter limited how many UID slots could be claimed in a single block on a subnet. That cap shaped burst registration behavior separately from the dynamic burn band and from network-wide rate limits (Subnet Hyperparameters, Glossary: Register).
It did not set registration price, immunity rules, or subnet capacity. It only bounded how many successful registrations could land in one block when that per-block cap model was active.
Current Registration Economics
Official documentation states that neuron registration is burn-based and always open on ordinary subnets, with admission governed by continuous burn pricing rather than per-block registration caps. Readers encountering max registrations per block in older material should treat it as legacy vocabulary and verify live behavior through the current burn hyperparameters on the relevant netuid (Subnet Hyperparameters, Understanding Neurons, Miner registration).
Documented Type and Default
The Subnet Hyperparameters
reference lists MaxRegistrationsPerBlock as a u16 with a default of one and marks the permission
required to set it as root
(Subnet Hyperparameters).
The constants reference exposes the startup default as
api.consts.subtensorModule.initialMaxRegistrationsPerBlock, listed as 1. That value matches the
documented one-registration default for the legacy per-block cap, while the current hyperparameter
documentation still marks the setting as unused for live neuron registration
(Subtensor Constants: initialMaxRegistrationsPerBlock,
Subnet Hyperparameters).
Distinction from Network Rate Limit
Network rate limit is a root-set block-counted wait between subnet-market creations. Max registrations per block named a per-block cap on neuron registrations within one subnet under older models (Subnet Hyperparameters).
- Network rate limit — wait between creating new subnet markets.
- Max registrations per block — legacy per-block neuron registration cap vocabulary.
Distinction from Registration Allowed
Registration allowed is the Boolean that decides whether a subnet accepts new neuron registrations at all. Max registrations per block named a per-block ceiling on how many registrations could succeed in one block under older models, and documentation now marks that cap as unused for live admission (Subnet Hyperparameters).
- Registration allowed — whether new registration is currently open on the subnet.
- Max registrations per block — legacy per-block registration cap vocabulary.
Distinction from Min Burn
Min burn sets the lower bound on the dynamic TAO burn required to register a neuron when registration is open. Max registrations per block did not set registration price; it only bounded how many successful registrations could land in one block when that per-block cap model was active (Subnet Hyperparameters).
Live admission is now governed by continuous burn pricing through burn half life and burn increase mult rather than per-block registration caps (Subnet Hyperparameters).
- Min burn — floor on the registration burn amount when entry is open.
- Max registrations per block — legacy per-block registration cap vocabulary.
Distinction from Max Allowed UIDs
Max allowed UIDs sets the ceiling on how many neurons a subnet may hold. Max registrations per block named how many UID slots could be claimed in a single block, not the total roster capacity (Subnet Hyperparameters).
- Max allowed UIDs — total neuron capacity ceiling on the subnet.
- Max registrations per block — legacy per-block registration burst cap vocabulary.