Blogs

In-Depth Analysis of OCS: Optical-Layer Direct-Connect Switching Technology

In-Depth Analysis of OCS: Optical-Layer Direct-Connect Switching Technology

In-depth analysis of OCS (Optical Circuit Switching) in AI training and high-performance computing (HPC) data centers, exploring its optical-layer direct-connect architecture, low-latency and high-bandwidth advantages, as well as its potential and limitations in complementing traditional electrical switching networks and optimizing large-scale collective communications.

What Is an XPO Transceiver? How Does It Differ from CPO?

What Is an XPO Transceiver? How Does It Differ from CPO?

What is an XPO (eXtra-dense Pluggable Optics) transceiver? Learn its architecture, key innovations such as dual-PCB design and liquid cooling, and how it compares with CPO for AI data center networks.

What Is MPO Trunk Cable? Structure, Types, and Application Scenarios Explained

What Is MPO Trunk Cable? Structure, Types, and Application Scenarios Explained

MPO Trunk cable is the backbone of high-density data center cabling. This guide covers what MPO Trunk is, how it differs from MPO jumpers and harnesses, four key specs, and deployment scenarios — from AI clusters to telecom backbone networks. Ideal for network engineers and procurement teams evaluating pre-terminated fiber solutions.

NADDOD × DGX Spark × OpenClaw: A Practical Guide to Local AI Agent Cluster Deployment

NADDOD × DGX Spark × OpenClaw: A Practical Guide to Local AI Agent Cluster Deployment

Learn how to deploy OpenClaw on NVIDIA DGX Spark with NADDOD's high-performance network solutions. A practical guide to building a secure, scalable local AI agent cluster for enterprises.

NVIDIA MGX Ecosystem: Building Modular Infrastructure for AI Factories

NVIDIA MGX Ecosystem: Building Modular Infrastructure for AI Factories

Explore the NVIDIA MGX ecosystem unveiled at GTC 2026, from Vera Rubin Pod to third-generation rack architecture. Learn how modular design, liquid cooling, and system-level co-design enable scalable AI infrastructure for training and inference.

NVIDIA Groq 3 LPX: A Low-Latency Inference Accelerator Designed for the NVIDIA Vera Rubin Platform

NVIDIA Groq 3 LPX: A Low-Latency Inference Accelerator Designed for the NVIDIA Vera Rubin Platform

NVIDIA Groq 3 LPX is a low-latency inference accelerator for the Vera Rubin platform. It adopts a GPU+LPU heterogeneous architecture, optimizes the decoding performance of large models, and achieves high throughput and predictable low latency in long context and high-concurrency scenarios, thus helping the development of intelligent agent systems and next-generation AI applications.

NVIDIA BlueField-4 STX Storage Architecture: Designed for an AI-Native Storage and Data Platform

NVIDIA BlueField-4 STX Storage Architecture: Designed for an AI-Native Storage and Data Platform

The NVIDIA BlueField-4 STX architecture enables high-performance, low-latency data access from GPUs through modular rack design and CMX external context storage, facilitating intelligent agent AI, multimodal large model inference and training, and promoting the deployment of scalable infrastructure for native AI data platforms.

Deep Dive into NVIDIA Groq 3 LPU: A New Choice for AI Inference

Deep Dive into NVIDIA Groq 3 LPU: A New Choice for AI Inference

NVIDIA Groq 3 LPU, integrated into the Vera Rubin platform, works with Rubin GPUs to accelerate low-latency, token-based AI inference with predictable performance and scalable multi-chip execution.

Exploring CPU、GPU、TPU and NPU: Architecture and Key Differences Introduction

Exploring CPU、GPU、TPU and NPU: Architecture and Key Differences Introduction

Understand the differences between CPU, GPU, TPU, and NPU in AI computing. Explore their architectures, performance characteristics, and ideal use cases for training and inference.