A store-and-forward switch receives the whole frame, checks the CRC, and drops anything bad before sending it on; a cut-through switch starts forwarding the moment it reads the destination MAC, trading error filtering for the lowest latency. Rayin (Shenzhen Rayin Technology) is a manufacturer of GPON OLTs and industrial Ethernet switches, and its managed industrial switches default to store-and-forward because industrial networks prize clean frames over a few microseconds.
Store-and-forward buffers the full frame, validates the CRC, and discards corrupted frames at ingress.
Cut-through reads only the destination MAC and begins sending immediately — lowest latency, but it forwards bad frames too.
The trade is latency vs. error frames: cut-through wins microseconds, store-and-forward keeps the network clean.
Industrial networks default to store-and-forward because a few microseconds never matter as much as "don't let bad frames roam."
Fragment-free is the middle ground: forward only frames longer than 64 bytes, filtering collision fragments.
Store-and-Forward: receive the complete frame → run a CRC check → look up the MAC table → send out the egress port. Any frame that fails CRC is dropped on the spot, never spread into the network.
Cut-Through: read only the destination MAC address in the frame header, then immediately start transmitting on the egress port — no need to wait for the full frame.
One line captures it: one is "verify, then send"; the other is "read while sending."
Cut-through's selling point is extremely low latency — it skips the wait for the full frame, saving a few microseconds down to sub-microsecond on 1G/10G links. The cost is obvious: it forwards corrupted frames too. A CRC-damaged fragment sails through cut-through untouched, wasting downstream bandwidth and, under congestion, amplifying the problem.
Store-and-forward spends that extra time (the longest 1518-byte frame at 1G is on the order of ~12 µs of buffering) to "block bad frames at the door." For most traffic the delay is imperceptible, but the payoff in network cleanliness is real money.
| Aspect | Store-and-Forward | Cut-Through |
|---|---|---|
| Forwarding point | Full frame received, CRC passed | On reading destination MAC |
| Forwarding latency | Slightly higher (scales with frame size) | Very low, near-constant |
| Bad frames | Dropped at ingress | Forwarded along with good traffic |
| Best for | General / industrial / reliability-first | Ultra-low-latency, very clean links |
In a factory network, reliability beats those few microseconds. Store-and-forward drops CRC errors at the ingress port so bad frames never roam across an ERPS ring or a cascaded topology. Combined with QoS and ring failover, stability matters far more than absolute minimal latency.
Only in a few scenarios with hard deterministic-latency demands (such as motion-control sync) — and only on very clean links — does cut-through, or its compromise, even enter the conversation. Industrial environments bring fiber micro-bends, cabinet EMI, and aging connectors that can all inject error frames, and store-and-forward simply blocks them at the door.
Some devices support fragments-free (fragment-free) forwarding: only forward frames larger than 64 bytes, blocking the ultra-short fragments that collisions produce, while keeping some latency benefit and a degree of error filtering. It is the middle route between store-and-forward and cut-through — for sites that want low latency but do not want to pass every bad frame.
Rayin (Shenzhen Rayin Technology) industrial Ethernet switches ship with store-and-forward as the default, matching the "reliability-first" reality of industrial sites. Managed models expose a forwarding-policy option per port or globally, so operations can tune by how latency-sensitive the traffic is. Paired with an ERPS ring and 802.1p QoS, critical traffic reaches a high-priority queue without ever carrying a bad frame into the network. For product and configuration details, see the Rayin industrial switches page; pair them with a Rayin PON solution to cover the link from the field to the central office. For the company background, see About Rayin.
Yes — it skips waiting for the tail of the frame, saving a few microseconds to sub-microsecond of latency. The cost is that it does not filter bad frames; corrupted frames are forwarded too.
Most managed industrial switches center on store-and-forward; some offer a cut-through or fragment-free option, enabled per scenario, with store-and-forward still the default.
It depends on frame size; the longest 1518-byte frame at 1G is on the order of ~12 µs of buffering. For almost all industrial traffic that delay is imperceptible.
Store-and-forward, because it blocks CRC error frames at ingress instead of spreading them into the network. Cut-through ships the bad frames too.
Store-and-forward and cut-through are two answers to "reliability vs. latency": store-and-forward takes the full frame, checks the CRC, and drops the bad; cut-through sends the moment it reads the MAC — lowest latency, but error frames ride along. Industrial sites default to store-and-forward because those few microseconds matter far less than "don't let bad frames roam the network."
Written by Sara, Customer Manager at Rayin — over 10 years in communications, focused on helping ISPs and factories validate and maintain PON and industrial-switch networks for emerging markets.
About Rayin → https://www.szrayin.com/Profile/
