Reach is one of the first things printed on an optical transceiver datasheet and one of the last things many buyers actually scrutinize. Yet the choice between short-reach (SR), DR4, and FR4 variants determines whether a link works at all, how much it costs to cable, and how cleanly a network can be broken out or upgraded later. Order the wrong reach and the module is either over-specified — paying for distance and optics the topology never uses — or under-specified, forcing a re-order weeks into a build. This guide frames reach as a procurement decision, not just an optical one, so buyers can match the module to the link on the first pass.
What "reach" actually encodes
In modern 400G, 800G, and 1.6T naming, the letters after the data rate describe a physical-layer specification: the modulation, the number of lanes, the fiber type, and the guaranteed distance. Three families dominate AI-cluster procurement:
- SR (Short Reach) — multimode optics for the shortest links, typically inside a rack or between adjacent racks. Distances are measured in tens of meters, and the transceiver is paired with multimode fiber (OM3/OM4/OM5).
- DR (Data-center Reach) — single-mode optics engineered for the intra-building distances typical of a spine-leaf fabric. "DR4" denotes four parallel single-mode lanes, each carried on its own fiber pair, with a nominal 500 m reach.
- FR (Far Reach) — single-mode optics for the longer runs that cross a large hall or connect adjacent buildings on a campus. "FR4" multiplexes four wavelengths onto a single fiber pair (duplex), with a nominal 2 km reach.
The distinction that trips up buyers most often is lane count versus wavelength count. DR4 uses parallel single-mode fiber — eight fibers (four pairs) in an MPO connector — while FR4 uses wavelength-division multiplexing to run four colors over a single duplex LC pair. That difference cascades into cabling, breakout, and cost long after the module is chosen.
The comparison that matters at order time
The table below summarizes the three families as a buyer would weigh them. Exact figures vary by data rate and generation; treat these as planning defaults and confirm against the specific product datasheet.
| Attribute | SR | DR4 | FR4 |
|---|---|---|---|
| Fiber type | Multimode (OM3/OM4) | Single-mode, parallel | Single-mode, duplex |
| Nominal reach | Tens of meters | ~500 m | ~2 km |
| Connector | MPO (multimode) | MPO-12 / MPO-16 | Duplex LC |
| Native breakout | Yes (to SR lanes) | Yes (to single-lane DR) | No (duplex link) |
| Typical role | In-rack / adjacent rack | Spine-leaf within a hall | Cross-hall / campus |
| Fiber-count sensitivity | High | High (parallel fibers) | Low (single pair) |
Two patterns fall out of this table. First, DR4 is the workhorse of AI spine-leaf fabrics because 500 m comfortably covers most links inside a single data hall while keeping the option to break out into four single-lane connections. Second, FR4 trades breakout flexibility for fiber economy — one duplex pair instead of eight fibers — which becomes decisive when structured cabling between halls is the constraint.
Breakout: where reach choice pays off or backfires
Breakout cabling is the reason reach type deserves careful attention rather than a default. A parallel single-mode module such as an 800G DR8 or a 400G DR4 can be broken out into multiple lower-rate links — for example, one 800G port fanning out to eight 100G destinations, or one 400G port to four 100G — because each lane already rides its own fiber. That makes DR-class optics attractive for connecting a high-radix switch to many endpoints without a separate breakout layer.
Duplex WDM optics such as FR4 do not break out the same way: the four wavelengths are multiplexed onto one pair and are recovered only by a matching module at the far end. Buyers who assume every module can fan out are sometimes surprised to find an FR4 link is point-to-point by design. The practical rule: if your topology depends on breakout, specify a parallel (DR-class) module and plan the MPO cabling accordingly. If the link is a straight point-to-point run over longer distance, FR4's single-pair economy usually wins. Our 400G to 800G migration guide covers how breakout topologies evolve as fabrics move up in speed.
A five-step checklist for specifying reach
- Measure the longest real link, not the average. A fabric is specified to its worst-case run. If a handful of links cross a hall at 700 m, an all-DR4 order (500 m) will leave those links short — plan FR4 or a longer-reach variant for the outliers rather than re-ordering later.
- Confirm the installed fiber plant. SR needs multimode; DR4 and FR4 need single-mode. Mixing a single-mode module onto multimode fiber (or vice versa) is a common and avoidable field failure. Verify the fiber type and connector (MPO vs LC) before the order, not on the loading dock.
- Decide breakout up front. If any link must fan out to lower-rate endpoints, choose a parallel DR-class module and size the MPO harnesses and patch panels for the lane count.
- Match connectors and polarity end to end. MPO-12 versus MPO-16, and the polarity/pinning convention, must be consistent across module, trunk, and patch. A reach spec that is correct but connector-mismatched still will not link.
- Leave headroom for the next generation. Reach families carry forward, but lane rates change. Buyers planning a 200G-per-lane future should confirm the fiber plant and connector choices they lock in today will carry an 800G or 1.6T upgrade. See our 1.6T OSFP readiness checklist for the forward view.
Common mismatches and how to avoid them
A few recurring procurement errors are worth naming directly:
- Buying FR4 for reach when DR4 would do. FR4 optics cost more per port than DR4 and add no value on a 300 m intra-hall link. Reach headroom is not free — specify to the link, not to the longest link you can imagine.
- Buying SR to save money on a link that will grow. Multimode reach shrinks as data rate rises. A multimode link that works today may fall short after a speed upgrade, whereas single-mode DR-class fiber tends to carry forward.
- Overlooking connector and polarity details. The single most common cause of a "dead" link that is optically fine is a connector or polarity mismatch in the structured cabling — a cabling problem wearing a transceiver problem's clothes.
When in doubt, the reach decision is best made alongside the cabling plan rather than after it. A supplier that can supply the module, confirm the matching MPO or LC assemblies, and hold consistent stock across a multi-phase build removes most of the risk from this decision. PhoScale supports SR, DR, and FR reach variants across 400G, 800G, and 1.6T on the products page, with documentation available on request and stocking arrangements described under supply assurance.
Conclusion
Reach is not a specification to accept by default. SR, DR4, and FR4 encode different fiber types, connectors, distances, and — critically — breakout behavior, and the right choice depends as much on your cabling plan and topology as on raw distance. Measure the worst-case link, confirm the fiber plant, decide breakout early, and match connectors end to end. Buyers who want a reach recommendation mapped to a specific fabric can submit an RFQ with link distances and topology, and receive a per-port reach plan alongside stock and lead-time information.