Both are pluggable transceiver form factors that share a nearly identical physical footprint. But under the hood, they're built for very different bandwidth tiers. Choosing wrong here isn't a minor mistake — it affects your capital budget, power and thermal planning, and whether your existing switches can even use the module.
This guide breaks down data rate, modulation, power draw, connectors, and real-world use cases so you can make the right call for your network.
TL;DR
- QSFP28 = 100G (4×25G NRZ); QSFP56 = 200G (4×50G PAM4)
- Both share the same mechanical footprint, but QSFP56 needs host support for PAM4 signaling
- QSFP56 draws more power (about 4.5–6.5W vs. ~3.5W) and costs more upfront
- Choose QSFP28 for mature 100G builds; QSFP56 for 200G and AI/HPC bandwidth
- Tested, warranty-backed refurbished modules cut upgrade costs on either form factor
QSFP56 vs QSFP28: Quick Comparison
| Spec | QSFP56 | QSFP28 |
|---|---|---|
| Data rate | 200G per port | 100G per port |
| Lanes & modulation | 4 × 50G PAM4 | 4 × 25G NRZ (100GAUI-4) |
| Short-reach connector | MPO-12 (SR4) | MPO-12 (SR4) |
| Longer-reach connector | Duplex LC | Duplex LC |
| Relative module cost | Higher (PAM4 DSP, stronger FEC) | Lower (mature NRZ) |
Throughput & signaling
QSFP56 delivers 200G per port over four 50G PAM4 electrical lanes. Arista notes the "56" refers to the maximum electrical interface rate per lane, not eight lanes as some assume.
QSFP28 delivers 100G per port (half the throughput) over four 25G NRZ lanes. That simpler signaling scheme has been the industry workhorse since 802.3bm-2015.
Power consumption
Vendor-published maximums vary by module type and reach:
| Module | Max Power |
|---|---|
| QSFP28 SR4/PSM4/CWDM4 (Cisco) | 3.5W |
| QSFP56 SR4 (Cisco) | 4.5W |
| QSFP56 FR4 (Cisco) | 5.0W |
| QSFP56 SR4 (Arista) | 4.5W |
| QSFP56 FR4 (Arista) | 6.5W |
These are maximums, not typical draw. Always confirm with the exact part number before finalizing a thermal budget.
Form factor, cost & compatibility
Both use MPO-12 for short-reach multimode (SR4) and duplex LC for longer single-mode variants. SNIA's SFF-8665 confirms QSFP28 and QSFP56 share a common mechanical ecosystem, so both fit the same physical slot on many switches.
QSFP56 modules cost more due to PAM4 DSP chips and stronger forward error correction. Mechanical fit still doesn't guarantee functional support—always check your switch's compatibility matrix before ordering.
What is QSFP28?
QSFP28 (Quad Small Form-factor Pluggable 28) is the 100G Ethernet standard built on 4x25Gbps NRZ lanes. It's been the backbone of enterprise and cloud data centers for nearly a decade.
Why it's stuck around:
- Mature, multi-vendor ecosystem with broad interoperability
- Lower power draw than newer PAM4-based modules
- Wide variety of reach options to match different topologies
Reach varies significantly by variant, based on Cisco's current 100G datasheet:
- SR4: 100 meters over OM4 multimode fiber, MPO-12 connector
- PSM4: 500 meters over single-mode fiber, MPO-12
- CWDM4: 2 km over single-mode fiber, duplex LC
- LR4: 10 km over single-mode fiber, duplex LC
Juniper's JNP-QSFP-100G-LR4 is a 10 km single-mode optic built for data center interconnects and long-haul 100G uplinks on QFX Series switches. On the shorter end, the JNP-QSFP-100G-SR4 handles 100-meter parallel multimode links for high-speed intra-data-center connections.
Use Cases of QSFP28
QSFP28 shows up everywhere 100G is still the target speed:
- Leaf-spine interconnects in traditional data center fabrics
- Aggregation layers connecting access switches to the core
- 100G breakout to 4x25G server links, common on switches like the Juniper QFX5200-32C
You'll find it heavily deployed across enterprise data centers, cloud hosting providers, and campus backbones. Juniper's QFX5110-32Q-AFI, for example, pairs four 100GbE QSFP28 uplinks with 40GbE access ports in a classic oversubscribed campus fabric design. That mix keeps QSFP28 a practical building block in production networks today.
What is QSFP56?
QSFP56 (Quad Small Form-factor Pluggable 56) doubles QSFP28's bandwidth to 200G, using four 50Gbps PAM4 lanes, all inside the same mechanical footprint as its predecessor.
Core benefits:
- Higher throughput per port without a new chassis form factor
- Fewer ports and cables needed for equivalent total bandwidth
- A practical migration bridge toward 400G as your fabric grows
Cisco's 200G QSFP56 datasheet lists two verified variants:
- SR4: 100 m over OM4 multimode (MPO-12, 4.5W max)
- FR4: 2 km over duplex single-mode fiber (duplex LC, 5W max)
Other labels like DR4 and LR4 exist in the market, but reach and power specs differ by vendor and part number. Don't assume they're interchangeable without checking the datasheet.
Use Cases of QSFP56
QSFP56 is built for environments where 100G ports are becoming the bottleneck:
- 200G spine and backbone links in hyperscale fabrics
- AI/HPC cluster interconnects, where GPU-to-GPU traffic demands more bandwidth per link
- High-density aggregation where port count is a physical constraint
Dell'Oro reported that 200/400G port shipments nearly doubled in 2023, with particularly strong growth among Tier 2/3 cloud providers. That's not a QSFP56-only figure (it blends 200G and 400G), but it's the clearest signal available that bandwidth-hungry environments are moving past 100G fast.
QSFP56 vs QSFP28: What is Better?
There's no universal winner. It comes down to three factors:
- Required bandwidth — do your applications actually need 200G, or is 100G still plenty?
- Host support — does your switch ASIC support PAM4 signaling and 50G-per-lane SerDes?
- Power and thermal budget — can your rack handle the extra 1–3 watts per port at scale?
Pick based on what matters most:
- QSFP28 if cost efficiency, simplicity, and longer native reach options matter most
- QSFP56 if higher per-port bandwidth and future-proofing outweigh the added cost and complexity
Either way, you don't need OEM new-unit pricing. Tested, warrantied refurbished QSFP28 and QSFP56 modules close the gap on both sides of this decision. Terabit Systems stocks Juniper and other major-brand optics, each run through a 5-step QA process and backed by a 1-year warranty.
Real World Example: Cost-Effective Network Upgrades
Picture a Tier-2 cloud provider running a 100G leaf-spine fabric that's hit capacity. Customer growth demands a jump to 200G spine links, but OEM pricing for new QSFP56 modules and switches is straining the capital budget. This is a common bind for providers competing against hyperscalers on price.
Instead of paying full OEM rates, the provider sources certified refurbished switches and QSFP56-compatible transceivers through a secondary-market reseller.
What that decision looks like in practice:
- Every module passes a 5-step quality control and testing process before sale
- Equipment ships with a 1-year warranty, so the provider isn't gambling on used gear
- Capital that would've gone to new hardware gets redirected toward the network build itself
Gartner's research on used-hardware resellers found buyers typically save 50%–60% on OEM hardware and support contracts, with savings reaching 85% on high-density contracts. It isn't an optics-only figure, but it shows how much providers can save by looking past new-unit pricing.
Growing providers don't need to overspend to scale bandwidth. If you're planning a QSFP28-to-QSFP56 migration, Terabit Systems' inventory of tested, warranty-backed hardware from Arista, Juniper, Cisco, and Extreme is worth a look before you sign an OEM quote.
Conclusion
QSFP28 remains the reliable, cost-effective choice for 100G networks that don't need more bandwidth yet. QSFP56 is the practical path forward when your traffic demands 200G per port. The right call depends on your host switch's support, your budget, and where your network is headed over the next few years.
Whichever form factor fits your roadmap, Terabit Systems supplies rigorously tested QSFP28 and QSFP56 optics—refurbished and legacy stock backed by a 1-year warranty—so you can scale bandwidth without overspending on new hardware.
Frequently Asked Questions
Is QSFP56 backwards compatible with QSFP28?
Mechanically, often yes, since both share a common footprint. Functionally, it depends on whether your host switch or NIC supports PAM4 and 50G-per-lane signaling. Always check the vendor's compatibility matrix first.
Can QSFP28 support 100G?
Yes. QSFP28 is purpose-built for 100G Ethernet using 4x25Gbps NRZ lanes, and it remains the industry standard for that speed tier.
Is QSFP-DD the same as QSFP56-DD?
No. QSFP-DD is a distinct double-density form factor with eight electrical lanes, supporting up to 400G or 800G. Standard QSFP56 uses four lanes for 200G: different architecture entirely, even though the names sound similar.
What does QSFP28 stand for?
Quad Small Form-factor Pluggable: four lanes, with "28" referring to the roughly 28Gbps-class signaling rate per lane used for 100G.
Is QSFP56 more expensive than QSFP28?
Generally, yes. PAM4 DSP chips, stronger FEC, and higher-speed components add cost. Refurbished QSFP56 options can narrow that price gap considerably.
Can I mix QSFP28 and QSFP56 modules in the same switch?
Many QSFP56-capable switches support QSFP28 modules for backward compatibility. The reverse usually isn't possible. QSFP28-only ports lack the PAM4 hardware needed to run QSFP56 optics.
