SFP Transceiver Compatibility: Form Factors, Reach, and Vendor Coding
Form factor vs speed, SR/LR/ER reach over multimode and single-mode, why switches reject uncoded optics, and when a DAC beats a transceiver.
Most "link won't come up" tickets blamed on optics are not hardware failures. They are mismatches: wrong form factor for the cage, wrong fiber for the optic, or a module the switch refuses to recognize. This guide covers the four checks that decide whether an SFP-family transceiver will actually work.
Form factor vs. speed: two separate questions
SFP, SFP+, and SFP28 modules share one physical cage; QSFP+ and QSFP28 share a larger one. The name tells you the electrical lane rate, not just the shape:
- SFP — 1GbE (and 100M variants), single lane
- SFP+ — 10GbE, same cage as SFP
- SFP28 — 25GbE, same cage again
- QSFP+ — 40GbE as four 10G lanes, larger cage
- QSFP28 — 100GbE as four 25G lanes
Down-negotiation usually works one generation back: most SFP28 ports run SFP+ modules at 10G, and most QSFP28 ports run QSFP+ at 40G. Support for 1G SFPs in SFP+ ports is the least consistent case — plenty of 10G ports do not support it at all. The reverse never works: a 25G module seats fine in an SFP+ cage but will not link. Verify against the switch datasheet, not the cage.
SR, LR, ER: matching reach to fiber
The reach suffix and the fiber plant must agree at both ends of the link.
| Type | Wavelength | Fiber | Typical reach |
|---|---|---|---|
| SR / SR4 | 850 nm | Multimode (OM3/OM4) | 10G: ~300 m OM3 / 400 m OM4; 25G lanes: ~70 m / 100 m |
| LR / LR4 | 1310 nm | Single-mode (OS2) | 10 km |
| ER | 1550 nm | Single-mode (OS2) | 40 km |
Parallel optics (SR4, PSM4) use MPO connectors and eight fibers rather than a duplex LC pair, so the QSFP choice changes your structured cabling, not just the module. Note that SR reach drops as lane rate rises — the 300 m you had at 10G on OM3 becomes roughly 70 m per lane at 25G.
Vendor coding: why switches reject working optics
Every module carries an EEPROM holding a vendor name, part number, and serial. The switch reads it at insertion and compares it against an approved list. Behavior on a mismatch ranges from a logged warning to an err-disabled port, and Cisco platforms are among the strictest — which is why third-party compatibility questions come up most often around Catalyst and Nexus gear.
Third-party modules coded for the target platform present the expected identity and behave like OEM parts, including DOM/DDM telemetry. On many Cisco platforms the service unsupported-transceiver command relaxes enforcement, but running unrecognized optics in production is the wrong fix when correctly coded modules cost a fraction of OEM list price. We stock coded SFP, SFP+, SFP28, and QSFP transceivers for current and end-of-life switch generations, matched to the platform they plug into.
Coding is per-vendor: a Cisco-coded optic is not automatically accepted by a Juniper or Arista switch, even when the silicon inside is identical.
DAC and AOC: skip the optics on short runs
Inside a rack or between adjacent racks, transceivers plus fiber are often the wrong answer. A passive DAC (direct-attach copper) covers runs up to roughly 5 m with no lasers, lower power draw, and two fewer optics in the failure chain. AOCs (active optical cables) extend that to tens of meters and take far less space in dense trays. Both come in breakout versions — one QSFP28 to four SFP28, for example — the standard way to fan a 100G port out to 25G server NICs. Browse DACs, AOCs, and fiber patch cables by length and coding.
The catch: a DAC carries vendor coding at both ends of one fixed cable. That is exactly the wrong shape for a mixed-vendor link.
Speccing a mixed-vendor link
When a server NIC from one vendor uplinks to a data center switch from another — the normal case in most environments — spec it this way:
- Same standard at both ends: 10GBASE-LR talks to 10GBASE-LR regardless of whose label is on each module.
- Each module coded for the device it plugs into. The two ends do not need matching coding; the fiber between them is vendor-neutral.
- Wavelength and fiber type consistent end to end, including patch panels.
- For short mixed-vendor runs, prefer two coded optics over a single-coded DAC, or use a DAC coded for both platforms where one exists.
Not sure which coding or reach variant your platform needs? Send us the switch or NIC model and the part number you are trying to match — compatibility matching by our sales team is free, every transceiver carries a 3-year warranty, and in-stock modules ordered before 12 PM EST ship the same day.
Frequently Asked Questions
Do third-party SFP transceivers work in Cisco switches?
Yes, provided the module is coded for Cisco. The switch reads the module's EEPROM at insertion; a module presenting the expected vendor identity links up and reports DOM data normally. Uncoded or wrongly coded optics trigger warnings or an err-disabled port depending on platform.
Can I use an SFP+ module in an SFP28 port?
Usually. Most SFP28 ports negotiate down to 10G with an SFP+ module, and most QSFP28 ports accept QSFP+ at 40G. The reverse never works — a 25G module in an SFP+ cage will not link. Confirm support in the switch datasheet.
What is the difference between SR and LR optics?
SR (short reach) runs at 850 nm over multimode fiber — roughly 300 m at 10G on OM3, less per lane at 25G. LR (long reach) runs at 1310 nm over single-mode fiber to 10 km. They are not interchangeable; wavelength and fiber type must match at both ends.
When is a DAC better than two transceivers?
For in-rack and adjacent-rack runs up to about 5 m, a passive DAC is cheaper, draws less power, and removes two optics from the failure chain. Beyond that, use an AOC or coded optics over fiber — and note a DAC carries coding on both ends, which matters on mixed-vendor links.