TL;DR
- ▸Clock synchronisation protocols form a precision hierarchy: NTP at milliseconds, PTP at microseconds and tens of nanoseconds, White Rabbit at sub-nanosecond, and emerging optical clock distribution beyond.
- ▸Each tier corresponds to a different cost, complexity and use case. Most production networks use more than one tier across different parts of the infrastructure.
- ▸The precision floor is steadily moving downward as use cases like AI training, quantum networking and HFT venue parity push tighter coordination requirements.
The precision hierarchy
Clock synchronisation protocols don't compete with each other; they sit at different rungs of a precision ladder. At the top of the ladder is NTP — millisecond-class accuracy, free, runs everywhere, the right answer for the vast majority of computer systems where rough time is enough. Below that is PTP (IEEE 1588) — sub-microsecond accuracy on hardware-aware networks, the standard for finance, broadcast, 5G fronthaul and any application where milliseconds aren't tight enough. Below that is White Rabbit — sub-nanosecond accuracy on fibre-distributed networks, originally developed at CERN for accelerator physics and now moving into commercial finance, AI and quantum applications.
Each rung represents a different cost, operational complexity and deployment model. Climbing the ladder is expensive, both in capital and in operational discipline. Most production networks live at one or two rungs and use the cheapest tier that meets their actual requirements.
What separates the tiers
Three engineering decisions drive the precision tier a protocol can deliver. First, where timestamps are captured — software (NTP) versus hardware (PTP, White Rabbit). Software timestamps include interrupt latency and kernel jitter that swamp microsecond-level measurements; hardware timestamps eliminate that variability. Second, whether the network compensates for asymmetric path delay — PTP boundary clocks and White Rabbit fibre paths both do; NTP doesn't. Third, the underlying frequency reference — NTP can run on any quartz oscillator; PTP grandmasters need disciplined OCXO/Rubidium; White Rabbit uses fibre-distributed phase locking back to a master reference.
These three decisions compound. A protocol that gets all three right delivers an order of magnitude better precision than one that gets one wrong. They're also the three places where deployments fail in practice — not because the protocol is incorrectly specified, but because the deployment skipped one of the engineering layers the protocol depends on.
Where the floor is heading
The precision hierarchy isn't static. Each rung has been pushed downward over the past decade as use cases have demanded tighter coordination. NTP got hardware timestamping support on modern NICs, pushing achievable accuracy from milliseconds toward low microseconds. PTP got Class C boundary clock specifications and multi-band multi-constellation GNSS, pushing achievable accuracy from hundreds of nanoseconds toward tens. White Rabbit went from a CERN-only physics tool to a commercial product line shipping into finance and AI training clusters.
Beyond White Rabbit, optical lattice clocks at the metrology end of the spectrum are now achieving precision good enough to detect gravitational time dilation across a one-metre vertical separation on Earth. As optical clock technology matures and moves out of national metrology labs, the bottom of the precision hierarchy will move down with it. The use cases that need White Rabbit today will need something more precise still in a decade.
How to think about your own deployment
The right question is not "which protocol is best?" but "which precision tier does my application actually need?" If you can't name a downstream consumer of your timestamps that would produce a worse outcome with millisecond-class precision, you don't need PTP. If you can't name an application that depends on sub-nanosecond temporal coordination, you don't need White Rabbit. Climbing the precision ladder costs real money and operational overhead, and the cost is wasted if the use case doesn't require it.
Equally, if your application has crossed a precision threshold and you're still running the cheaper tier, the failures will be subtle and persistent — drift you can't explain, audit trails that don't quite line up, distributed systems that can't agree on event ordering. These are the symptoms of being one tier too low for the actual requirement, and the fix is to climb the ladder.
Frequently asked questions
What is the precision hierarchy of clock synchronisation protocols?+
Why does hardware timestamping matter?+
What is White Rabbit?+
How do I choose which precision tier I need?+
Related reading
Blog · Protocols
Precision Time Protocol vs NTP: When Each Belongs in Production
The honest engineering comparison between Precision Time Protocol and NTP — what each protocol can actually deliver, where the boundary lives, and how to choose between them without falling for either side's marketing.
Blog · White Rabbit
Why White Rabbit is Changing the Game for Finance, AI and Quantum Innovation
White Rabbit — the sub-nanosecond fibre-distributed timing protocol developed at CERN — is moving out of physics labs and into financial trading floors, AI training clusters and quantum networking testbeds. What it actually does, and why precision below one nanosecond is now a commercial requirement.
Blog · Protocols
Sync Showdown: NTP vs PTP vs TSN vs EtherCAT
Four protocols, four different precision tiers, four different deployment contexts. A comparative guide to NTP, PTP, TSN and EtherCAT — what each one does, when each one is correct, and how to choose without falling for any side's marketing.