#NCCL

45 related articles

Jul 23, 2026·3 min

Qwen 3.8 Open Source Incoming, DeepSeek Coding Agent, and the Latest in China's AI Ecosystem

Qwen 3.8 open source incoming, DeepSeek Resonics native coding agent debuts, Kimi pauses new subscriptions amid compute pressure, and Qiyimole shows domestic GPU interconnect solution at WAIC.

Jul 23, 2026·7 min

Alibaba Open-Sources 2.4T Qwen Model as Chinese AI Sweeps the Global Competitive Landscape

Alibaba open-sources a 2.4 trillion parameter Qwen model and launches the Qwen Token Plan. Chinese models surge, Kimi K3 tops global rankings, and China's AI is reshaping the global competitive landscape.

Jul 17, 2026·2 min

vLLM Inference Framework Deep Dive: A Practical Guide to High-Concurrency DeepSeek Deployment

A deep dive into vLLM's five core features — KV cache management, continuous batching, and distributed inference — plus a full setup guide for deploying DeepSeek on a cloud server.

Jul 16, 2026·3 min

LongCat 2.0: Meituan Trains a 1.6 Trillion Parameter LLM on Domestic Chips

Meituan open-sources LongCat 2.0, a 1.6T-parameter MoE model trained on 50,000+ custom chips without NVIDIA GPUs or Google TPUs, rivaling OpenAI and Google.

Jul 12, 2026·3 min

Nvidia's Dilemma: How the Commoditization of the Compute Market Is Turning Against Its Creator

Nvidia's revenue keeps rising, yet its stock has fallen ~15% from its peak. A deep dive into the double-edged sword of compute commoditization—when cloud giants build their own chips, how much of Nvidia's pricing power and moat remains?

Jul 12, 2026·3 min

SlickToken: A Deep Dive into GPU Fleet and Agentic Workflow Planning

SlickToken is a GPU fleet and agentic workflow planning tool for AI teams, supporting offline simulation, load testing, and capacity planning—no internet required to protect enterprise data.

Jul 12, 2026·3 min

DeepSeek-V4 Flash RL Training Successfully Ported to AMD MI355X GPU: A Complete Breakdown

The Miles team and AMD announce the full port of DeepSeek-V4 Flash RL training to AMD Instinct MI355X GPUs on ROCm, boosting AIME pass@1 from 0.39 to 0.49—a milestone for compute ecosystem diversity.

Jul 12, 2026·6 min

DeepSeek Builds Its Own AI Chips: The Strategic Logic and Challenges Behind Compute Independence

DeepSeek is entering AI chip development, targeting compute autonomy. This article analyzes its motivations, software-hardware synergy, chip R&D challenges, and impact on China's AI vertical integration.

Jul 11, 2026·7 min

Practical RL Compute Optimization: Core Strategies to Boost CPU and GPU Utilization

An in-depth look at why CPU and GPU utilization is low in RL training, covering vectorized environment parallelism, distributed Actor-Learner architectures, GPU-side simulation (Isaac Gym/Brax), and Ray RLlib practice.

Jul 11, 2026·3 min

The Software Compounding Effect: The Real Deciding Factor in AI Infrastructure Competition

Beyond the GPU arms race, software is becoming AI infrastructure's true compounding engine. A deep dive into optimization stacking, the CUDA ecosystem moat, and how open source scales individual breakthroughs into community wealth.

Jul 11, 2026·2 min

AI Agents for SGLang Kernel Optimization: Real-World Performance Gains and Engineering Methodology

SGLang's team converted expert knowledge into agent skills, achieving 71.4% throughput gains, TTFT reduced from 456ms to 168ms. A deep dive into agent-assisted kernel optimization methodology.

Jul 10, 2026·6 min

Tencent Hunyuan HY3 Official Version Open-Sourced: The Strongest Under 300B, at 1 Yuan per Million Tokens

Tencent Hunyuan HY3 official version is open-sourced under Apache 2.0, priced as low as 1 yuan per million input tokens, with major gains in agents, reasoning, coding, and long context. On the same day, Meituan open-sourced its trillion-parameter LongCat 2.0.

Jul 10, 2026·5 min

TensorRT Multi-Device Inference: Engineering Practices for Breaking Through Single-GPU Memory Bottlenecks

NVIDIA TensorRT now supports multi-device inference via pipeline and tensor parallelism, distributing large models across multiple GPUs to break through single-card memory limits.

Jul 10, 2026·7 min

Reverse-Engineering CUDA-checkpoint: The Technical Principles and Practice of Cracking GPU Cold Starts

An in-depth analysis of reverse-engineering Nvidia CUDA-checkpoint to accelerate GPU cold starts. Covers checkpoint/restore, Serverless GPU prospects, and VRAM snapshot challenges.

Jul 9, 2026·6 min

AMD Ryzen AI Halo Dev Kit: A $4,000 Local LLM Powerhouse Taking on Apple and NVIDIA

AMD officially unveils the Ryzen AI Halo local AI dev kit, priced around $4,000 with 128GB unified memory, capable of running 70B LLMs locally. An in-depth look at its specs, pricing, and market competition.

Jul 7, 2026·5 min

Nonuniform Tensor Parallelism: NVIDIA's New Fault-Tolerance Approach to Boost LLM Training Goodput

NVIDIA introduces Nonuniform Tensor Parallelism, letting GPUs bear different compute loads so training can continue without checkpoint rollback during hardware failures—boosting LLM training Goodput and fault tolerance at scale.

Jul 7, 2026·6 min

AMD MI355X Running GLM5.2: Inference Costs at Half of NVIDIA Blackwell

AMD MI355X achieves 2,626 tokens/sec throughput running GLM5.2 at half the TCO of NVIDIA Blackwell. Deep analysis of the technical logic, ROCm ecosystem progress, and market implications.

Jul 6, 2026·6 min

A 284B Model Running on a MacBook: The Real Breakthrough and Barriers of Local DeepSeek Inference

Redis creator runs 284B-parameter DeepSeek model on a MacBook Pro at 26 tokens/sec using a pure C engine, asymmetric quantization, and MoE architecture.

Jul 3, 2026·5 min

Reproducing GitHub Projects from Scratch: A Practical Step-by-Step Guide

A complete guide to reproducing GitHub projects from scratch: due diligence, virtual environments, dependency installation, script interpretation, and breakpoint debugging — helping grad students and junior developers run others' code efficiently.

Jun 23, 2026·3 min

DAQIRI Platform Explained: Deep Integration of High-Speed Data Acquisition and Real-Time AI Inference

Deep dive into how the DAQIRI platform embeds NVIDIA GPU-accelerated computing into high-speed data acquisition pipelines, enabling real-time AI inference for industrial inspection, scientific experiments, and autonomous driving.