200 G - QSFP56
QSFP56 — 200G. 200G PAM4 — four lanes at 50G each, mostly seen in HPC and storage.
Overview
Quad SFP56 — same mechanical envelope as QSFP28 (SFF-8665), but four lanes of 50G PAM4 instead of 25G NRZ. Most production 200G deployments are NVIDIA Spectrum + Mellanox ConnectX-6 InfiniBand-on-Ethernet (RoCE) clusters, and storage fabric where 100G is too slow but 400G is overkill.
Key specifications
| Speed | 200G |
|---|---|
| Electrical interface | 4 lanes × 26.5625 Gbaud PAM4 (50G per lane) |
| Signalling | PAM4 with KP4 / RS-FEC |
| FEC | RS-FEC (544,514) required |
| Typical power draw | 4.5 W – 7 W per module |
| Connectors | MPO-12 (SR4, DR4), LC duplex (FR4, LR4) |
| MSA / standards | SFF-8665 (mechanical); CMIS 4.0 / 5.0 (management) |
| Introduced | 2018 (IEEE 802.3cd) |
Available variants
| Part number | Standard | Wavelength | Fiber | Reach | Notes |
|---|---|---|---|---|---|
NAP-QSFP56-SR4 |
200GBASE-SR4 | 850 nm | MM OM4 | 100 m | MPO-12 PAM4 |
NAP-QSFP56-DR4 |
200GBASE-DR4 | 1310 nm | SM parallel | 500 m | MPO-12 |
NAP-QSFP56-FR4 |
200GBASE-FR4 | LAN-WDM | SM | 2 km | LC duplex |
NAP-QSFP56-LR4 |
200GBASE-LR4 | LAN-WDM | SM | 10 km | LC duplex |
NAP-QSFP56-AOC |
200G AOC | 850 nm | — | 3–30 m | Common in HPC racks |
NAP-QSFP56-DAC |
200G DAC | — | Twinax | 1–3 m | Passive |
Coded for these platforms
Every NetAPI QSFP56 module is OEM-coded for the target switch and validated against current production firmware. Common targets:
- NVIDIA Spectrum-3 (SN4000 series)
- Mellanox ConnectX-6 (HDR InfiniBand, 200G Ethernet)
- Arista 7060X4, 7280R3
- Cisco Nexus 9300-GX (200G via 100G splits + aggregation)
Recommended for
- InfiniBand HDR (200 Gbps) on NVIDIA fabric
- 200G RoCE storage interconnect
- High-performance computing clusters between 100G and 400G generations
Engineering notes
- PAM4 signalling at 50G/lane has roughly 9 dB less SNR margin than NRZ at equivalent symbol rates — clean fiber and low-loss connectors matter more here.
- 200G AOC is the dominant interconnect in HDR InfiniBand racks for NVIDIA SuperPOD-class deployments.
Need a coding or reach you don't see? NetAPI custom-codes against the full Cisco, Juniper, Arista, Dell, HPE, Aruba, Palo Alto, Huawei, and NVIDIA OUI sets. Contact us with your switch model and we'll quote.