NADDOD-Building an Intelligent World with Everything Connected

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 ,...

In today's high-performance networking market, the Broadcom Thor2 and NVIDIA CX7 are two important network interface cards (NICs). Each has its own benefits in terms of performance, energy efficiency,and cost-effectiveness, etc. Thor2 is known for its low power consumption and multi-port configuration capabilities, while the CX7 is gaining attention for its RDMA technology and high performance acceleration. The different features of Thor2 and CX7 NICs  also leads to differences in application scenarios. Thor2 is more suited for large-scale data centers, with a particular focus on high throughput and flexible scalability. CX7 is primarily targeted at high-performance computing, AI, and deep-learning applications, optimized for low-latency and high-throughput requirements.   Although these 2 are different in many aspects, they both demonstrate strong technical advantages in their own fields. No doubt that they are both great choices for powering AI/ML workloads ...

Since the unveiling of xAI Colossus , the world’s largest AI supercomputing cluster, it draw widely attention to Ethernet networking. Unlike most supercomputers that utilize technologies such as InfiniBand, Ethernet was chosen for this project. Colossue was built using the GPU giant's Spectrum-X Ethernet fabric.   This choice is highly strategic. Ethernet, the protocol that underpins the Internet, offers superior scalability. Many enterprises are already deeply familiar with Ethernet, relying on it as a trusted networking standard. However, for a long time, they lacked a solution to adequately support the characteristics of AI workloads using the Ethernet protocol.   To solve this problem, here came the NVIDIA Spectrum-X. As the world’s first high-performance Ethernet fabric, it represent a significant leap forward, ensuring that Ethernet remains a robust and future-proof technology in an era of exponential data growth.   What is Spectrum-X? Spectrum-X, the world's leadin...

  NVIDIA GB200 Interconnect Architecture Analysis-NVLink, InfiniBand, and Future Trends 1.   NVLink Bandwidth Calculation There is a lot of confusion in NVIDIA's calculation of NVLink transmission bandwidth and the concept of SubLink/Port/Lane. Typically, the NVLink 5 bandwidth of a single GB200 is 1.8TB/s, which is calculated by bandwidth-oriented individuals in bytes per second (Byte/s). However, on NVLink Switches or IB/Ethernet switches and network cards, Mellanox's perspective calculates it in terms of network bandwidth, usually in bits per second (bit/s), based on the transmitted data bits.   Here, we'll explain in detail the calculation method of NVLink. Starting from NVLink 3.0, it is composed of four differential pairs to form a "sub-link" (often referred to by NVIDIA as Port/Link, with some ambiguity in definition). These 4 pairs of differential signal lines simultaneously contain both the receiving and transmitting directions of signal lines. Typically,...