C-LIGHT telephone TEL:+86 132 6656 7067    
Language
C-LIGHT search

SFP28 vs QSFP28: 25G Access and 100G Aggregation Interface Selection

By C-LIGHT Marketing 丨 Dec 23, 2025
Table of Contents

    SFP28 and QSFP28 serve distinct roles in a network. SFP28 carries a single 25Gbps lane and is used for server access, Top-of-Rack connections, and 5G fronthaul. QSFP28 aggregates four 25Gbps lanes into 100Gbps and is used for uplinks, aggregation, leaf-spine fabric, and 4x25G breakout. The two form factors are physically incompatible, and the selection should be driven by the port type, lane architecture, breakout requirements, power budget, and the network layer being designed.

    1. What Is SFP28?

    SFP28 is a single-lane 25Gbps optical module form factor built on the SFP+ physical footprint. It carries one 25Gbps electrical lane using NRZ modulation and is widely deployed for 25G server access, Top-of-Rack switching, enterprise aggregation, storage networking, and 5G fronthaul[reference:0].

    SFP28 modules typically consume around 1.5W to 2W depending on reach and optics type, with short-reach and AOC variants often below 1W and longer-reach single-mode variants approaching 2W[reference:1][reference:2]. The form factor is backward compatible with SFP+ cages at 10G, but an SFP+-only port generally cannot operate an SFP28 module at 25G.

    In 5G networks, SFP28 is the workhorse for eCPRI and O-RAN fronthaul over 25GbE, providing the symmetric uplink and downlink capacity required for centralized and cloud RAN architectures[reference:3].

    2. What Is QSFP28?

    QSFP28 stands for Quad Small Form-factor Pluggable 28. It combines four 25Gbps lanes in a single module to deliver 100Gbps aggregate bandwidth. QSFP28 is the mainstream 100G interface for switch uplinks, aggregation layers, leaf-spine fabric, and high-density 100G links in cloud and enterprise data centers[reference:4][reference:5].

    QSFP28 modules typically consume around 3W to 5W depending on reach and optics type, with short-reach variants near 3.5W and longer-reach ER-class optics approaching 5W[reference:6][reference:7]. While the per-module power is higher than SFP28, a single 100G QSFP28 link can replace four 25G SFP28 links at the uplink level, which often improves power efficiency per 100G of bandwidth.

    A defining capability of QSFP28 is breakout support. A single 100G QSFP28 port can be split into four independent 25G interfaces using a breakout cable and compatible SFP28 modules on the downstream end. This 100G-to-4x25G breakout is common in leaf-spine architectures, where spine switches use QSFP28 ports to connect to multiple leaf switches with 25G server access[reference:8][reference:9].

    3. Key Specifications

    ItemSFP28QSFP28
    Lane architecture1 × 25Gbps lane4 × 25Gbps lanes
    Aggregate speed25Gbps100Gbps
    ModulationNRZNRZ per lane (typical for 100G Ethernet)
    Physical sizeApprox. 13.4mm × 56.5mmApprox. 18.4mm × 73.4mm
    Cage compatibilityPhysically incompatible with QSFP28 cagesPhysically incompatible with SFP28 cages
    Typical powerRoughly 1.5–2WRoughly 3–5W per module
    Breakout supportNot applicable as a single-lane form factorSupports 4 × 25G breakout when the switch and cabling support it
    Common optical connectorsLC duplex for most opticsMPO/MTP for SR4/PSM4; LC duplex for CWDM4, LR4, ER4
    Representative reachSR up to 70m OM3 / 100m OM4; LR up to 10kmSR4 up to 70m OM3 / 100m OM4; CWDM4 up to 2km; LR4 up to 10km; ER4 up to 40km
    Primary network roleServer access, ToR, edge links, 5G fronthaulUplinks, aggregation, leaf-spine fabric, high-density 100G

    4. Physical Incompatibility and Breakout Adapters

    Important distinction: SFP28 and QSFP28 are not physically interchangeable. The cage dimensions and pin definitions differ, so an SFP28 module cannot be inserted into a QSFP28 port, and a QSFP28 module cannot be inserted into an SFP28 port[reference:10]. The common way to connect SFP28 links to a QSFP28 port is through a breakout cable or a dedicated QSFP28-to-SFP28 adapter that maps the SFP28 lane to lane 1 of the QSFP28 connector[reference:11]. This is an architectural connection method, not direct module interoperability.

    5. Power, Density, and Cost Perspective

    SFP28

    • Lower per-port power, which simplifies thermal design in high-density 25G access platforms.

    • Mature 25G NRZ ecosystem with broad SR, LR, ER, DWDM, DAC, and AOC availability.

    • Cost-effective for server access and fronthaul ports where 25G per port matches the traffic profile.

    QSFP28

    • Higher per-module power, but one 100G module can replace four 25G links at the uplink level, improving efficiency per 100G of bandwidth.

    • Higher front-panel 100G density than four discrete SFP28 ports for the same aggregate throughput.

    • Breakout capability allows a 100G spine port to serve four 25G leaf connections, simplifying cabling and port utilization.

    A published comparison notes that a direct QSFP28 100G link can be more cost-effective and significantly more power-efficient than four separate SFP28 25G links for the same 100G aggregate, with reported figures around 5.5W for QSFP28 versus roughly 14W for 4×SFP28[reference:12]. Actual values depend on the specific modules, reach, and platform design.

    6. Application Fit

    6.1 SFP28 for 25G Access and Fronthaul

    SFP28 is the default 25G interface for server access, Top-of-Rack switching, storage networking, virtualized clusters, and 5G fronthaul. In a data center, servers connect to leaf switches through SFP28 ports at 25G. In 5G, SFP28 carries eCPRI and O-RAN fronthaul traffic over 25GbE, supporting the bandwidth and latency requirements of centralized and cloud RAN[reference:13].

    6.2 QSFP28 for 100G Aggregation and Leaf-Spine

    QSFP28 is used for switch uplinks, aggregation layers, and spine-leaf fabric. In a typical leaf-spine design, leaf switches provide 25G SFP28 server access and use QSFP28 uplinks at 100G to connect to spine switches. Spine switches use QSFP28 ports for 100G interconnect and can break out those ports into 4 × 25G to connect multiple leaf switches efficiently[reference:14][reference:15].

    6.3 Breakout and Mixed Design

    QSFP28 breakout allows one 100G port to serve four 25G downstream connections through a breakout cable and SFP28 modules at the far end. This is common in leaf-spine fabrics and in racks where a 100G uplink is distributed to multiple 25G server ports. The switch must support breakout mode, and the cabling and optics must be matched to the breakout configuration.

    7. Selection Guidance

    Start with the port type on the switch or NIC and the required per-port speed. For 25G server access, storage, or fronthaul, SFP28 provides a low-power, cost-effective solution with a mature optics portfolio. For 100G uplinks, aggregation, or spine interconnect, QSFP28 provides the required bandwidth and can support 4 × 25G breakout to connect downstream 25G devices.

    Verify whether the switch port supports breakout, the correct breakout cable or adapter is available, and the far-end optics are compatible. Do not attempt to insert an SFP28 module into a QSFP28 cage or vice versa. Use the switch datasheet, optics compatibility list, and breakout configuration guide as the basis for selection.

    8. Frequently Asked Questions

    Q1: Can I plug an SFP28 module into a QSFP28 port?

    No. SFP28 and QSFP28 use different cage sizes and pin definitions. Direct insertion is not possible. To connect an SFP28 link to a QSFP28 port, use a dedicated QSFP28-to-SFP28 adapter or a breakout cable designed for that purpose.

    Q2: Can a QSFP28 port support 25G?

    Yes, when the switch and port support breakout mode. A 100G QSFP28 port can be split into four 25G interfaces using a breakout cable and SFP28 optics on the downstream end. Not all QSFP28 ports support breakout, so check the switch specification.

    Q3: Is QSFP28 always more power-hungry than SFP28?

    Per module, yes, QSFP28 typically consumes more power (around 3–5W versus 1.5–2W for SFP28). However, for a 100G uplink, one QSFP28 module can replace four SFP28 modules, so the power per 100G of bandwidth can be lower with QSFP28 depending on the comparison.

    Q4: Which one is used for 5G fronthaul?

    SFP28 is the common 25G interface for 5G eCPRI and O-RAN fronthaul. QSFP28 is used at aggregation and transport layers where 100G uplinks are required, but the fronthaul radio link itself typically uses SFP28 at 25G.

    Q5: What is the main advantage of QSFP28 breakout?

    Breakout allows a single 100G switch port to serve four 25G downstream connections. This improves port utilization and can simplify cabling in leaf-spine fabrics where a spine port needs to connect to multiple 25G leaf or server ports.

    9. Conclusion

    SFP28 and QSFP28 are complementary interfaces at different levels of the network. SFP28 provides single-lane 25G connectivity for server access, ToR, storage, and 5G fronthaul. QSFP28 aggregates four 25G lanes into 100G for uplinks, aggregation, spine-leaf fabric, and breakout applications. They are not physically interchangeable, but they can be connected through breakout cables and adapters. The correct choice depends on the port type, required bandwidth, breakout needs, power budget, and network layer.

    Designing 25G access and 100G aggregation layers?

    Evaluate SFP28 and QSFP28 modules based on port compatibility, breakout support, power budget, and reach requirements.

    Explore C-LIGHT optical modules for 25G and 100G network connectivity

    For any questions, please contact us by email or WhatsApp.

    Email: sales@c-light.com

    WhatsApp: +86 132 6656 7067

    Related Articles

    Call
    Top