With AI and data centers expanding, the demand for high-speed optical modules is experiencing unprecedented growth. At the same time, the pace of optical module upgrades to higher speeds is accelerating in order to cope with the continuously climbing data processing and transmission requirements resulting from model training. While the previous transition from 100G to 400G took more than three years, the generation replacement from 800G to 1.6T is expected to be shortened to less than two years. In 2024, 1.6T transceivers have already appeared, and it is estimated that in 2025, the production of 1.6T optical modules may grow rapidly.
To cope with the growing demand for AI and the rising demand for computing, NADDOD has launched the OSFP-1.6T-2xDR4, a high-speed silicon phtonics transceiver specifically tailored for 1.6T InfiniBand XDR. InfiniBand XDR(eXtended Data Rate)is the new speed generation in the IBTA roadmap after NDR, providing 2x increased bandwidth over the previous generation to meet the rapid performance and scalability demands of AI and HPC data centers. With SerDes support of 200Gb/s per lane, XDR-capable adapters and switches will provide unprecedent 800Gb/s speeds per port, and support for XDR switch-to-switch connections at 1.6Tb/s.
Designed for 1.6T InfiniBand XDR, OSFP-1.6T-2xDR4 is a cost-effective, high-performance module with 2x800Gb/s dual-port OSFPs and 8x212Gb/s PAM4 signal modulation for optical and electrical measurement. By pairing it with two MPO-12/APC fiber patch cables, it can achieve up to 500 meters of data transmission. This silicon-based optical transceiver can achieve high-speed interconnection in areas such as AI data centers and cloud computing due to its ultra-high transmission rate and reliability.
Currently, the OSFP-1.6T-2xDR4 module is primarily used for interconnection between NVIDIA Quantum-X800 Q3400 switches, building a platform that supports 1.6T InfiniBand network switches, especially designed for trillion-parameter AI models. The NVIDIA Quantum-X800 Q3400 switch is the world’s first switch to adopt single-channel 200Gb/s technology, offering 144 ports with a connection speed of 800Gb/s. All these ports require 72 pieces 1.6T OSFP optical modules to achieve full functionality.
Key Features of OSFP-1.6T-2xDR4
The industry's first 200G-PAM4 optical DSP—Marvell 5nm DSP
- Strong FEC Error Correction: It enables NADDOD 1.6T transceiver to guarantee the accuracy of data transmission and significantly reduce the error rate, thus greatly improving the reliability and efficiency of AI model training and ensuring the accuracy of the training results.
- Low Latency: With the superior performance of Marvell's 5nm DSP chip, OSFP-1.6T-2xDR4 can effectively reduce data transmission latency, which is critical for AI applications with high real-time requirements, making sure that the AI system is able to acquire and process the latest information in a timely manner, so as to make accurate real-time decisions.
- High-speed data processing: Marvell 5nm DSP chips accelerate data transmission and processing, significantly reducing the time required for AI model training and inference, and improving overall work efficiency.
Self-developed single-wave 200G silicon optical chip
The 1.6T SiPh module uses NADDOD's self-developed silicon optical chip, with a high-speed operation of 200G per channel to ensure rapid transmission of AI massive data, significantly reducing the time required for AI model training and inference. It guarantees the stability of AI clusters in long-time and high-intensity data transmission tasks.
Advantages
1.6T transceiver, with its exceptional high-bandwidth capability, plays a critical role in accelerating large-scale model training and data processing for AI clusters while enabling a smooth transition to higher-performance infrastructure. In network expansion scenarios, the 1.6T transceiver demonstrates superior resource efficiency. For instance, achieving 3.2Tbps bandwidth with traditional 800G modules requires four modules and four links, whereas the 1.6T transceiver accomplishes the same with only half the modules. This not only significantly reduces deployment costs but also enhances network scalability, providing robust support for future intelligent computing.
In Summary
NADDOD has been focusing on high-performance optical networking solutions since its establishment, and has always emphasized the research of new technologies and products in the industry. With years of rich experience in the AI field, we precisely grap the market trends and has taken the lead in launching 1.6T silicon optical modules to meet the demand for ultra-high bandwidth and low power consumption in ultra-large-scale data centers and high-performance computing. In terms of high-speed networking, NADDOD has a full line of products including optical transceivers, aoc/dac cables, NICs and switches, as well as corresponding technology reserves and solutions.
Feel free to contact us to learn about the advanced 1.6T SiPh optical module and explore how NADDOD’s comprehensive solutions meet the high-speed networking needs.


- 1NADDOD Delivers the First 400G OSFP-RHS SR8 Module to the Market
- 2Inside DeepSeek's 10,000 GPU Cluster: How to Balance Efficiency and Performance in Network Architecture
- 3NADDOD Leads in Compatibility and Performance on Broadcom Thor2 & NVIDIA CX7
- 44 Common Spectrum-X Product Solutions in Ethernet Networking
- 5Single-Phase vs. Two-Phase Immersion Cooling in Data Centers