{"id":17344,"date":"2025-05-23T10:00:46","date_gmt":"2025-05-23T10:00:46","guid":{"rendered":"https:\/\/www.fibermall.com\/blog\/?p=17344"},"modified":"2025-09-05T07:23:35","modified_gmt":"2025-09-05T07:23:35","slug":"ocp-emea-2025-fibermalls-1-6t-pluggable-optical-module-based-on-224g-per-lane","status":"publish","type":"post","link":"https:\/\/www.fibermall.com\/blog\/fibermall-1600g-module-based-on-224g-per-lane.htm","title":{"rendered":"OCP EMEA 2025: FiberMall\u2019s 1.6T Pluggable Optical Module Based on 224G per Lane"},"content":{"rendered":"\n

The rapid growth of artificial intelligence (AI) and machine learning has driven a substantial demand for higher bandwidth, making the 224G per lane configuration essential. Enhanced energy efficiency is equally critical in managing the ever-increasing power requirements of data centers.<\/p>\n\n\n\n

\"data<\/figure>\n\n\n\n

With copper-based transmission reaching its physical limits, optical communication has become indispensable for scalable AI deployment across servers and racks. Notably, IEEE 802.3dj is actively engaged in standardizing the 224G per lane technology, and numerous vendors demonstrated live 1.6T 224G per lane solutions at OFC2025.<\/p>\n\n\n\n

\"link<\/figure>\n\n\n\n

The latest revision of the OSFP MSA introduces an innovative chassis design engineered to address the escalating thermal challenges. The OSFP 2×1 cage design, as depicted in the accompanying illustration, permits the direct mounting of liquid cooling plates onto the module. This approach enhances thermal management and achieves a superior power-to-performance balance while supporting more densely packed AI server configurations.<\/p>\n\n\n\n

\"OSFP<\/figure>\n\n\n\n

For instance, FiberMall\u2019s 1.6T OSFP<\/a> 224 2*DR4 module exhibits an energy efficiency of 18.75\u202fpJ\/bit when utilizing a 5nm DSP. By transitioning to a 3nm DSP, the power consumption is reduced to 16.25\u202fpJ\/bit without sacrificing performance. An additional power reduction is realized via LPO, which achieves an impressive 6.25\u202fpJ\/bit, positioning this configuration as an ideal solution for energy-efficient AI scaling.<\/p>\n\n\n\n

\"1.6T<\/figure>\n\n\n\n

For the 1.6T OSFP224 2xDR4 module, comparative OMA and TDEcQ test results between the 3nm and 5nm DSP implementations have been obtained. Although the 3nm DSP remains in its early design stage, its initial performance already meets the specified criteria and trends towards the outcomes of the 5nm design.<\/p>\n\n\n\n

\"OMA<\/figure>\n\n\n\n

The 3nm DSP design satisfies the bit error rate thresholds across all channels with a considerable margin, confirming its compliance with specifications and production feasibility.<\/p>\n\n\n\n

\"BER<\/figure>\n\n\n\n

An end-to-end validation of the 1.6T OSFP224 2xDR4 LRO optical module demonstrated that the bit error rate (BER) in every channel is significantly lower than the 1E-4 threshold. This result attests to the robust signal integrity at a production level, marking a crucial milestone toward deploying a 1.6T system without DSPs while achieving energy savings.<\/p>\n\n\n\n

\"An<\/figure>\n\n\n\n

FiberMall is also actively developing and expanding application support for both the 1.6T OSFP224 2xVR4 and 1.6T OSFP224 2xFR4 modules. As illustrated by the accompanying eye diagrams, early design prototypes show remarkable maturity and link stability. The overarching objective is to ensure comprehensive coverage across various distances\u2014from short-range connections within AI clusters to long-distance spine-leaf interconnections.<\/p>\n\n\n\n

\"long-distance<\/figure>\n\n\n\n

In summary, the burgeoning field of artificial intelligence is propelling data centers to demand solutions that offer higher bandwidth and lower power consumption. The adoption of a 1.6T optical system<\/a> based on 224G per lane technology represents a pivotal advance for future AI infrastructure. With industry developments advancing rapidly, FiberMall\u2019s live demonstrations of these optical modules precede the finalization of the standards. The newly introduced OSFP cage design, with its liquid cooling support, facilitates superior thermal management. Moreover, the promising performance of the 3nm DSP, along with further reductions in power consumption via linear optical techniques and the proven reliability of the LRO module, collectively lay the groundwork for future DSP-free designs.<\/p>\n\n\n\n

Moving forward, considerations such as deployment scalability in diverse data center environments and the integration of these innovations with emerging network architectures may also offer exciting avenues for exploration.<\/p>\n\n\n\n

<\/p>\n