{"id":12524,"date":"2026-06-10T15:02:55","date_gmt":"2026-06-10T07:02:55","guid":{"rendered":"https:\/\/ascentoptics.com\/blog\/?p=12524"},"modified":"2026-06-10T15:02:55","modified_gmt":"2026-06-10T07:02:55","slug":"qsfp112-vs-qsfp28","status":"publish","type":"post","link":"https:\/\/ascentoptics.com\/blog\/qsfp112-vs-qsfp28\/","title":{"rendered":"QSFP112 vs QSFP28: Key Differences and Migration Guide"},"content":{"rendered":"

The same QSFP cage that carries your 100G traffic today can theoretically carry 400G tomorrow. The catch? The module inside must speak a completely different electrical language. That is the central tension network engineers face when comparing QSFP112<\/u><\/a>\u00a0vs QSFP28. Both optical transceiver modules share identical physical dimensions, but the leap from 100G to 400G is not a simple swap. It demands new switch ASICs, different modulation schemes, and a clear-eyed assessment of what infrastructure you can keep and what must change.<\/p>\n

 <\/p>\n

 <\/p>\n

What Is QSFP28? The 100G Standard<\/strong><\/h2>\n

Technical Foundation: 4 x 25G NRZ<\/strong><\/h3>\n

QSFP28<\/u><\/a>\u00a0(Quad Small Form-factor Pluggable 28) has been the dominant 100G optical transceiver module since its widespread adoption around 2016. It achieves 100G data transmission through four electrical lanes, each operating at 25 Gbps using NRZ (Non-Return-to-Zero) modulation. This straightforward lane architecture makes QSFP28 reliable, thermally efficient, and broadly compatible across networking equipment vendors.<\/p>\n

 <\/p>\n

\"100G<\/p>\n

 <\/p>\n

Common QSFP28 Module Types<\/strong><\/h3>\n

Network engineers choose QSFP28 variants based on transmission distance and fiber type:<\/p>\n\n\n\n\n\n\n\n\n
Module Type<\/b><\/strong><\/td>\nFiber<\/b><\/strong><\/td>\nReach<\/b><\/strong><\/td>\nUse Case<\/b><\/strong><\/td>\n<\/tr>\n
100GBASE-SR4<\/td>\nMMF (OM4)<\/td>\nUp to 100 m<\/td>\nIntra-rack and adjacent-rack links<\/td>\n<\/tr>\n
100GBASE-LR4<\/td>\nSMF<\/td>\nUp to 10 km<\/td>\nData center interconnect, metro links<\/td>\n<\/tr>\n
100GBASE-ER4<\/td>\nSMF<\/td>\nUp to 40 km<\/td>\nLong-haul telecom and carrier networks<\/td>\n<\/tr>\n
100GBASE-PSM4<\/td>\nSMF<\/td>\nUp to 2 km<\/td>\nSpine-leaf architectures, DCI<\/td>\n<\/tr>\n
100GBASE-CWDM4<\/td>\nSMF<\/td>\nUp to 2 km<\/td>\nCost-optimized data center links<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

<\/h3>\n

 <\/p>\n

Why QSFP28 Remains Relevant<\/strong><\/h3>\n

QSFP28<\/u><\/a>\u00a0modules typically draw 3.5 to 6 watts per unit, making them thermally manageable in standard enterprise racks. The ecosystem is mature, with broad vendor support from Cisco, Arista, Juniper, and dozens of white-box manufacturers. For networks where 100G bandwidth is sufficient, QSFP28 continues to deliver stable performance at a cost-per-port that 400G modules have not yet matched.<\/p>\n

 <\/p>\n

 <\/p>\n

What Is QSFP112? The 400G Evolution<\/strong><\/h2>\n

Technical Foundation: 4 x 112G PAM4<\/strong><\/h3>\n

QSFP112 is the 400G optical transceiver module that pushes the traditional 4-lane QSFP form factor to its architectural limit. Instead of scaling lane count, QSFP112 scales per-lane speed: four electrical lanes operating at approximately 112 Gbps using PAM4 (Pulse Amplitude Modulation 4-level) modulation. This delivers an aggregate 400G data transmission rate within the same physical envelope as QSFP28.<\/p>\n

 <\/p>\n

\"400G<\/p>\n

 <\/p>\n

How QSFP112 Achieves 400G in the Same Form Factor<\/strong><\/h3>\n

The key engineering decision behind QSFP112 was to preserve backward mechanical compatibility while maximizing bandwidth density. PAM4 modulation packs two bits per symbol, effectively doubling the data rate at the same baud rate compared to NRZ. However, PAM4 is far more sensitive to signal loss, noise, and PCB trace quality. QSFP112 modules, therefore, incorporate sophisticated Digital Signal Processing (DSP) chips to equalize and recover signals at 112G per lane.<\/p>\n

 <\/p>\n

Common QSFP112 Module Types<\/strong><\/h3>\n\n\n\n\n\n\n\n
Module Type<\/b><\/strong><\/td>\n\n

Fiber<\/b><\/strong><\/p>\n<\/td>\n

Reach<\/b><\/strong><\/td>\n\n

Use Case<\/b><\/strong><\/p>\n<\/td>\n<\/tr>\n

400GBASE-SR4<\/a><\/td>\n\n

MMF (OM4)<\/p>\n<\/td>\n

Up to 100 m<\/td>\n\n

Intra-rack 400G links<\/p>\n<\/td>\n<\/tr>\n

400GBASE-DR4<\/a><\/td>\n\n

SMF<\/p>\n<\/td>\n

500 m<\/td>\n\n

Leaf-spine data center fabrics<\/p>\n<\/td>\n<\/tr>\n

400GBASE-FR4<\/a><\/td>\n\n

SMF (duplex LC)<\/p>\n<\/td>\n

2 km<\/td>\n\n

Campus, aggregation, short DCI<\/p>\n<\/td>\n<\/tr>\n

400GBASE-LR4<\/a><\/td>\n\n

SMF (duplex LC)<\/p>\n<\/td>\n

10 km<\/td>\n\n

Metro and backbone links<\/p>\n<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

 <\/p>\n

 <\/p>\n

QSFP112 vs QSFP28: Head-to-Head Technical Comparison<\/strong><\/h2>\n

Understanding the differences between QSFP112 and QSFP28 requires looking beyond the form factor label. Here is how they stack up across the dimensions that matter for network design.<\/p>\n

 <\/p>\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Specification<\/b><\/strong><\/td>\nQSFP28<\/b><\/strong><\/td>\nQSFP112<\/b><\/strong><\/td>\n<\/tr>\n
Aggregate Data Rate<\/strong><\/td>\n100 Gbps<\/td>\n400 Gbps<\/td>\n<\/tr>\n
Electrical Lanes<\/strong><\/td>\n4 lanes<\/td>\n4 lanes<\/td>\n<\/tr>\n
Per-Lane Speed<\/strong><\/td>\n25 Gbps<\/td>\n~112 Gbps<\/td>\n<\/tr>\n
Modulation<\/strong><\/td>\nNRZ<\/td>\nPAM4<\/td>\n<\/tr>\n
Form Factor Dimensions<\/strong><\/td>\n18.4 x 89.4 x 8.5 mm<\/td>\n18.4 x 89.4 x 8.5 mm<\/td>\n<\/tr>\n
Typical Power Consumption<\/strong><\/td>\n3.5-6W<\/td>\n8-15W<\/td>\n<\/tr>\n
Signal Processing<\/strong><\/td>\nMinimal DSP<\/td>\nAdvanced DSP required<\/td>\n<\/tr>\n
Host ASIC Requirement<\/strong><\/td>\n25G NRZ SerDes<\/td>\n112G PAM4 SerDes<\/td>\n<\/tr>\n
Primary Use Case<\/strong><\/td>\nEnterprise 100G, mainstream DC<\/td>\n400G upgrades, AI\/HPC clusters<\/td>\n<\/tr>\n
Ecosystem Maturity<\/strong><\/td>\nMature (2016+)<\/td>\nEmerging (2021+)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

 <\/p>\n

Data Rate and Lane Architecture<\/strong><\/h3>\n

QSFP28 achieves 100G with four lanes at 25 Gbps each. QSFP112 quadruples that to 400G by pushing each lane to approximately 112 Gbps. The lane count stays the same, but the signal integrity challenges increase exponentially. Every connector, trace, and via in the host PCB must handle 112G PAM4 signaling, which is significantly more demanding than 25G NRZ.<\/p>\n

 <\/p>\n

Modulation: NRZ vs PAM4<\/strong><\/h3>\n

NRZ transmits one bit per symbol: a high voltage represents 1, a low voltage represents 0. It is simple, power-efficient, and tolerant of moderate signal degradation. PAM4 transmits two bits per symbol using four distinct voltage levels. This doubles throughput at the same baud rate but requires more complex receivers, tighter channel budgets, and higher power DSP. For network engineers, this means QSFP112 demands cleaner signal paths and better thermal management than QSFP28.<\/p>\n

 <\/p>\n

\"NRZ<\/p>\n

 <\/p>\n

Power Consumption and Thermal Design<\/strong><\/h3>\n

Here is where the upgrade math gets real. A QSFP28 module typically draws 3.5 to 6 watts. A QSFP112 module draws 8 to 15 watts depending on type and reach. At the switch level, a 32-port device fully populated with QSFP112 modules can consume 256 to 480 watts in optics alone, versus 112 to 192 watts with QSFP28.<\/p>\n

When Marcus Chen, a senior infrastructure engineer at a mid-size cloud provider, planned his first 400G spine layer, he assumed his existing thermal design would scale linearly. It did not. The jump from roughly 6W to 12W per module meant his standard 15 kW racks could handle fewer switches than planned. He ended up redistributing the spine layer across two additional racks, which affected his cabling topology and power distribution units.<\/p>\n

 <\/p>\n

Backward Compatibility: The Reality<\/strong><\/h3>\n

Many QSFP112-based platforms support QSFP56 and QSFP28 modules for lower-speed operation. However, backward compatibility depends on the switch ASIC, firmware, and vendor implementation, so compatibility should always be verified before deployment.\u00a0This is a significant advantage for phased upgrades.<\/p>\n

However, the reverse is not true. A legacy QSFP28 port supports only 25G NRZ signaling. You cannot plug a QSFP112 module into a QSFP28 port and expect it to function at 400G. The physical dimensions match, but the electrical interface does not. This is the single most common source of confusion during procurement.<\/p>\n

 <\/p>\n

 <\/p>\n

When to Use QSFP28 vs QSFP112<\/strong><\/h2>\n

Choose QSFP28 If…<\/strong><\/h3>\n