--- title: "OSFP vs. QSFP-DD for 800G: The Port Density Math Nobody Shows You" slug: "800g-osfp-vs-qsfp-dd-port-density" category: "Hardware Selection" tags: ["800G", "OSFP", "QSFP-DD", "port density", "switch selection", "thermal management"] seo_focus_keyword: "800G OSFP QSFP-DD comparison port density" word_count_target: 1200 difficulty: intermediate --- When 800G arrived in production, it brought with it a form factor argument that remains genuinely unresolved. OSFP (Octal Small Form-factor Pluggable) and QSFP-DD (Quad Small Form-factor Pluggable — Double Density) both deliver 800G, both have real products on the market, and both have partisans who will tell you the other one is a dead end. The truth is more useful than either camp admits: they serve different optimization targets, and choosing between them requires doing actual math rather than accepting vendor narrative. **The physical reality** OSFP is larger. That's the starting point. An OSFP module is approximately 22.58 mm wide and 107.8 mm deep. A QSFP-DD is 18.35 mm wide and 89.4 mm deep. The extra width of OSFP is not an accident — it provides more surface area for heat dissipation and space for the eight 100G electrical lanes that connect to the host ASIC. OSFP was designed with the thermal requirements of 800G optics as a primary constraint, not an afterthought. QSFP-DD is mechanically backwards-compatible with QSFP28 and QSFP+ cages, which is a significant installed base advantage. If you have 400G QSFP-DD infrastructure, you have the right cage geometry for 800G QSFP-DD modules — though the electrical and thermal specifications differ enough that you should not assume a cage designed for 400G QSFP-DD will simply handle 800G without validation. **Port density: the actual numbers** On a 1U switch with a standard 19-inch rack width, the front panel real estate is fixed. Let's work with a practical example: Arista's 7800R3 series and comparable Cisco Nexus 9000 platforms. A 1U switch supporting OSFP typically achieves 32 OSFP ports. At 800G per port, that's 25.6 Tbps of front-panel bandwidth. A 1U switch supporting QSFP-DD at the same 800G speeds can often achieve 36 or even 40 ports in a dense implementation — approximately 32 Tbps. That's a real 25% port density advantage for QSFP-DD, and it compounds in spine-layer deployments. If your spine tier runs 128 full-bisection uplinks to a leaf layer, the QSFP-DD spine switch requires fewer chassis units to deliver equivalent bandwidth. In a 3-tier fabric, the cumulative difference in rack units, power draws, and cabling can be significant. However, this density advantage evaporates under thermal load. The maximum power dissipation per OSFP module is specified at 15W for current 800G modules, with some optical variants approaching 20W. QSFP-DD 800G modules target a 14W maximum, but many real-world implementations sit at 12–13W due to the tighter thermal budget imposed by the smaller form factor. Push the cage to 100% utilization with high-power coherent or long-reach optics, and QSFP-DD switches frequently hit thermal throttling thresholds that OSFP switches handle without incident. **Thermal limits in practice** The critical number to understand is the per-cage thermal budget, not the per-module number. Switch ASIC vendors like Broadcom publish thermal design power (TDP) for the ASIC itself, but the cage management system — the combination of cage size, airflow path, and heat sink geometry — determines whether you can realistically run all ports at full rated optical power. For 800G SR8 short-reach optics (100m OM4 reach), module power consumption is typically 8–10W for OSFP and 7–9W for QSFP-DD. These are well within the thermal envelope of both form factors, and full port density is achievable. For 800G DR8 optics (500m SMF reach, parallel fiber), modules run 12–14W. Both form factors handle this, but you should verify cooling configuration. For 800G FR8 optics (2km SMF reach), some modules approach 18–20W. This is where OSFP's larger thermal mass becomes decisive. Running FR8 at full density in QSFP-DD is often not possible — vendors will explicitly rate those switches at 50–75% port utilization for high-power optical variants. For coherent 800G ZR modules (80–120km), power consumption hits 20–25W. These modules are available in OSFP form factor (the 800ZR OSFP modules from Acacia/Cisco and Coherent) but not realistically in QSFP-DD for production use. If coherent is in your application, OSFP is not optional. **When each makes sense: a practical decision guide** Choose QSFP-DD when your application is primarily short to medium reach (SR8, DR8, FR4 at 800G), your fabric is bandwidth-density limited rather than thermal-limited, and QSFP backward compatibility with existing 400G infrastructure matters. DCI and hyperscale intra-datacenter fabrics running parallel SMF at distances under 2km are the sweet spot. The higher port density genuinely reduces capex in these scenarios. Choose OSFP when you need coherent or extended-reach optics, when you're building a long-haul or metro aggregation layer that will run high-power DWDM modules, or when you expect optic generations to increase power consumption over the switch's service life. OSFP's larger thermal envelope is future insurance. The Arista 7130 series, Cisco 8000 series, and Juniper PTX series platforms all offer OSFP configurations for exactly this reason. There is also a practical consideration about vendor roadmap alignment. OSFP is the form factor preferred by most coherent transceiver manufacturers for next-generation 1.6T implementations. The 1.6T OSFP specification is further along than the 1.6T QSFP-DD equivalent, in part because the thermal headroom required for 1.6T coherent simply doesn't fit in the QSFP-DD envelope. If you're designing infrastructure with a 5–7 year operational life, and that life includes 1.6T, OSFP gives you a more credible upgrade path. **Breakout cabling and what it does to your math** Both form factors support breakout: an 800G OSFP or QSFP-DD port can break out to 8×100G or 2×400G using an appropriate breakout cable or cassette. This is where the density comparison becomes context-dependent. If you're using 800G ports as 2×400G breakouts for leaf-switch connectivity in a spine-leaf fabric, the QSFP-DD density advantage (more 800G ports = more breakout endpoints) can meaningfully increase your oversubscription headroom. If you're using breakout to 8×100G for server connectivity, the marginal density difference between OSFP and QSFP-DD per switch matters less than the cable management implications of running 8-fiber MPO breakout fans in a high-density rack. **The cable plant consideration nobody mentions** Both 800G DR8 and SR8 use parallel fiber — 16-fiber MPO16 connectors (or dual MPO8). This is a significant cable plant commitment. If your existing infrastructure uses duplex LC or MPO12, a migration to 800G at meaningful density requires rethinking your fiber trunk architecture, and neither OSFP nor QSFP-DD changes that. FR8 uses eight parallel SMF lanes in a dual-MPO configuration. Only FR4 (four wavelengths on two fibers) and coherent ZR maintain duplex LC compatibility, which is a strong argument for these form factors in campuses and enterprise metro rings where MPO infrastructure isn't already in place. The bottom line is simple enough: for dense intra-DC 800G fabrics without coherent requirements, QSFP-DD wins on density. For anything involving coherent optics, high-power extended-reach modules, or roadmapping toward 1.6T, OSFP is the right platform. Buy the switch for the optical application, not the form factor preference.