400G QSFP-DD: Choose SR8, DR4, FR4, LR4 or 400ZR

Which 400G QSFP-DD Module Should You Choose?

• Choose SR8 for short multimode links when MPO-16 cabling is available and the required 8×50G or 2×200G breakout mode is supported.
• Choose SR4.2 when an existing MPO-12 multimode plant must be preserved or a compatible 4×100G BiDi migration path is required.
• Choose DR4 for parallel single-mode links up to 500 m, especially when 4×100G-DR breakout is part of the design.
• Choose FR4 for native 400G point-to-point links over duplex single-mode fiber up to 2 km.
• Choose LR4-6 or LR4-10 for longer duplex single-mode links after confirming the exact IEEE or MSA specification.
• Choose 400ZR or OpenZR+ when the link is a coherent DWDM transport application rather than a conventional grey-optics Ethernet link.

400G QSFP-DD Module Comparison

For current vendor-specific reach, connector, breakout and maximum-power data, consult the Cisco 400G QSFP-DD module data sheet or the equivalent documentation for the platform being deployed.

What Does 400G QSFP-DD Actually Define?

QSFP-DD defines an eight-lane pluggable form factor and the host-side mechanical, electrical, thermal, pinout and management framework. It does not define a universal optical reach, connector or interoperability profile. Those properties come from IEEE Ethernet standards, OIF implementation agreements, other MSAs and vendor specifications.

The current QSFP-DD MSA specification archive is therefore the right source for the module and cage framework, while each optical PMD must be validated against its own specification. A broader QSFP-DD technical overview can help readers separate form-factor rules from optical-interface rules.

Host Electrical Lanes and Optical Lanes Are Different Layers

A 400G QSFP-DD host port may present eight electrical lanes at roughly 50G per lane. A DR4, FR4 or LR4 module can use internal gearbox or digital signal processing functions to map those host lanes into four optical lanes at roughly 100G per lane. A coherent module performs a much more complex conversion into a single modulated DWDM carrier.

This distinction is fundamental to breakout planning. A switch may support a logical 4×100G port mode while the installed optical module exposes a line-side architecture that cannot connect to the intended four endpoints.

Key Terms Used in This Guide

PMD

Physical Medium Dependent interface. It defines the optical signaling, wavelength or lane structure, reach and media requirements for the link.

FEC

Forward Error Correction. The host and remote endpoint must use compatible coding behavior so that the receiver can correct the expected pre-FEC errors.

CMIS

Common Management Interface Specification. It defines how a host identifies, configures and monitors many modern pluggable modules.

OSNR

Optical Signal-to-Noise Ratio. It is a central coherent-link metric because amplified noise and filtering can limit performance even when fiber distance appears acceptable.

Choose by Deployment Scenario

400GBASE-SR8: Short-Reach Parallel Multimode

400GBASE-SR8 uses eight transmit fibers and eight receive fibers, normally through an MPO-16 interface. It is designed for short data-center connections where multimode economics, high port density and straightforward parallel optics are more important than minimizing fiber count.

SR8 is a strong candidate when the route is within the qualified OM3 or OM4 reach, a correct MPO-16 plant is available, and the host explicitly supports the required native or breakout mode. It is not enough to confirm that the building has multimode cabling. An installed MPO-12 trunk does not provide the sixteen active positions required for a direct SR8 link.

Multimode grade also matters. The OM1-to-OM5 distance guide provides useful context for bandwidth and reach differences, but the final design must follow the transceiver specification rather than a generic fiber-distance chart.

SR8 Breakout Considerations

Commercial SR8 implementations commonly support selected modes such as 8×50G or 2×200G. Some platforms may offer other mappings, but no breakout mode should be treated as universal. Confirm the switch ASIC mode, operating-system release, module SKU, fan-out assembly and far-end optics as one tested system.

400GBASE-SR4.2: Reusing MPO-12 Multimode Infrastructure

SR4.2 uses four multimode fiber pairs, with two bidirectional wavelengths operating on each pair. This architecture enables 400G over an MPO-12 interface and can extend to 150 m on OM5 in compliant implementations.

Its main value is not simply reach. SR4.2 can preserve a brownfield MPO-12 plant that would otherwise need recabling for SR8. It can also align with four compatible 100G SR1.2 or BiDi endpoints when the host and module support that breakout mode.

The bidirectional optical design is different from SR8 even though both are multimode solutions. Readers evaluating the wavelength pairing and endpoint behavior can review this explanation of BiDi transceiver technology.

Check connector polish, key orientation and polarity carefully. A generic label such as "MPO-12 cable" does not fully define an assembly. The MTP versus MPO selection guide explains the terminology and mechanical considerations that should be recorded in the bill of materials.

400GBASE-DR4: Parallel Single-Mode and 4×100G Migration

DR4 carries four parallel 100G optical lanes over single-mode fiber for links up to 500 m. It normally uses an MPO-12 interface with four transmit and four receive fibers active.

The lane structure makes DR4 the most direct optical architecture for connecting one 400G host port to four compatible 100G-DR endpoints. However, "4×100G" is only a service-rate description. The four far-end modules must use a matching 100G optical PMD and compatible FEC behavior.

A DR4 module cannot be connected directly to arbitrary 100G-FR, 100G-LR, CWDM4 or SR4 modules merely because every endpoint is labeled 100G. A passive fan-out cable rearranges physical fibers; it does not translate wavelengths, modulation or coding.

Because DR4 depends on single-mode infrastructure, verify the actual cable grade rather than relying on jacket color. This OS1 versus OS2 single-mode fiber guide helps distinguish common plant types, but production acceptance should use labels, records and test results.

400GBASE-FR4: Duplex Single-Mode up to 2 km

FR4 carries four 100G-class CWDM wavelengths over one fiber in each direction, normally through a duplex LC interface. It is often the most practical starting point for a native 400G link over an existing duplex OS2 pair when the measured route and loss remain within the module specification.

Its principal infrastructure advantage is fiber efficiency: only one fiber pair is required. The tradeoff is that the wavelength multiplexing happens inside the module. A passive LC fan-out cannot split FR4 into four independent 100G clients.

When four lower-speed endpoints are required, use a purpose-built 4×100G parallel module or an active muxponder or transponder architecture. The difference between those device roles is explained in this guide to transceivers and transponders.

Why FR4 Is Often Evaluated Before LR4

For a route within 2 km, FR4 usually deserves the first engineering and commercial evaluation because it can meet the service without paying for reach that the link does not need. That is not a universal rule. LR4 may still be justified by the measured loss, receiver range, platform support, spare strategy, future route extension or procurement standardization.

400GBASE-LR4-6 and 400G-LR4-10: Specify the Exact Standard

The label "400G LR4" is not precise enough for a purchase specification. IEEE 400GBASE-LR4-6 defines a 6 km class interface. The separate 100G Lambda MSA 400G-LR4-10 specification defines a four-wavelength CWDM interface for duplex single-mode links from 2 m to at least 10 km.

Both families use four optical wavelengths over a duplex pair, but procurement teams should not assume identical optical budgets, receiver limits or multi-vendor interoperability. The request for quotation should state:

• the required IEEE or MSA designation;
• minimum and maximum channel insertion loss;
• minimum link distance or receiver-overload constraints, if applicable;
• approved host platforms and software releases;
• interoperability requirements at both ends;
• maximum module power and operating-temperature range.

Buying the longest-reach optic "to be safe" can increase cost, power and qualification work without improving the actual link. Select LR4 because the measured channel and service requirement justify it, not because its distance label is larger.

400ZR and OpenZR+: Coherent Optics for Engineered DWDM Links

400ZR is a coherent optical interface created for interoperable 400G Ethernet transport over data-center interconnect links. The current OIF implementation-agreement library lists the 400ZR specification revisions. The defined application includes amplified point-to-point DWDM links up to 120 km and unamplified links governed by an optical-loss budget.

A coherent link cannot be selected by distance alone. Engineering inputs include fiber attenuation, OSNR, mux and demux loss, amplifier noise, channel spacing, launch power, chromatic dispersion, ROADM filtering and host support. A route marked "80 km" may pass or fail depending on the complete optical path.

OpenZR+ expands the operating space with higher-gain FEC, multiple line rates, client multiplexing and modes intended for regional and long-haul applications. The OpenZR+ Specification Version 3.0 adds enhanced modes and broader application support.

Client multiplexing in a coherent module is not the same as passive breakout. It is an active mapping function implemented by the module and host. Verify the exact host interface, line mode, firmware and management support before assuming that a specific OpenZR+ module can aggregate the desired clients.

A Seven-Step 400G QSFP-DD Selection Workflow

This workflow replaces separate, repetitive selection checklists with one end-to-end process.

Step 1: Freeze the Service Definition

Write the required endpoint configuration before comparing modules: 1×400G to 1×400G, 1×400G to 4×100G, 1×400G to 2×200G or another explicit mapping. Identify both endpoint PMDs, not only their Ethernet rates.

Step 2: Verify the Host Platform

Record the chassis, line card, port, operating-system release, license requirements, approved module list, third-party optic policy, supported breakout modes and maximum module power. A QSFP-DD module may fit mechanically yet remain disabled, report limited diagnostics or fail to enter the requested port mode.

Step 3: Audit the Fiber Plant

Confirm fiber grade, measured length, connector type, polish, gender, keying, polarity, active fiber positions, patch panels, cassettes and tested insertion loss. Cable color is a convention, not proof.

Step 4: Calculate the Optical Budget

Add fiber attenuation, mated-connector loss, splice loss, cassette or patch-panel loss, mux or demux loss and an engineering margin. Compare the result with the channel insertion-loss limit and transmitter and receiver limits in the exact module datasheet.

Step 5: Select the Compatible Optical Architecture

• MPO-16 multimode, short reach: evaluate SR8.
• MPO-12 multimode with compatible BiDi support: evaluate SR4.2.
• MPO-12 parallel single-mode up to 500 m: evaluate DR4.
• Duplex single-mode up to 2 km: evaluate FR4.
• Duplex single-mode beyond 2 km: evaluate LR4-6 or LR4-10.
• Engineered DWDM transport: evaluate 400ZR or OpenZR+.

Step 6: Verify Breakout, FEC and CMIS Behavior

Confirm port mode, electrical lane mapping, line-side PMD, fan-out assembly, remote optics, FEC and module-management behavior. The OIF CMIS resource page is the authoritative starting point for current CMIS work, but platform software can still implement different revisions and feature subsets.

For parallel-optics deployments, use a documented assembly rather than a generic cable description. The MPO breakout cable selection guide covers fiber count, polarity and endpoint choices that should be frozen in the production BOM.

Step 7: Validate Before Volume Purchase

Test the exact production combination of switch, software, module SKU, cable plant and far-end optic. Record module recognition, link-up time, transmit and receive levels, pre-FEC performance, corrected and uncorrected errors, temperature, alarms, breakout stability, reboot behavior and software-upgrade behavior.

Worked Link-Budget Example: 1.8 km FR4 Candidate

The following calculation is illustrative. Replace every assumed value with the actual cable records, measured insertion loss and module datasheet limits.

If the selected FR4 datasheet permits 4 dB of channel insertion loss, this illustrative design leaves 0.87 dB of remaining margin. That is a candidate, not an approval. The receiver power limits, minimum link distance, reflectance, actual measured loss and host qualification must still pass.

For a more detailed explanation of the loss terms and measurement logic, review this guide to insertion loss in fiber networks.

Cost, Power and Upgrade Tradeoffs

Unit price alone rarely identifies the lowest-cost 400G architecture. Compare optics, cabling, installation, patching, spare inventory, host licensing, cooling and future migration as one system.

Use maximum power, not a typical marketing value, for chassis thermal design. The QSFP-DD MSA defines power classes, while the actual host may impose lower per-port limits, port-group restrictions or airflow derating.

Common Configuration Errors and Their Consequences

FAQ