#multi-GPU communication

65 related articles

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

GPUStack in Action: Complete Guide to Day 0 Deployment of DeepSeek-V4-Flash

Deploy DeepSeek-V4-Flash DSpark on 8× H20-141G using GPUStack's SGLang backend on Day 0. Full walkthrough of Web UI config, parameter tuning, and 200 tokens/sec benchmark results.

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 16, 2026·2 min

ExLlamaV3 v1.0.0 Official Release: Faster Inference, Lossless KV Quantization, and Streamlined Dependencies

ExLlamaV3 v1.0.0 releases with lossless KV cache quantization via kernel fusion, removal of flash-attention-2/xformers, major GEMM/GEMV gains, and broader tensor parallelism support.

Jul 16, 2026·3 min

AI Daily: Alibaba Open-Sources Dance Model, World Model Launches, First AI-Native Phone Debuts

Alibaba open-sources 14B dance model Wan-Dancer, AutoNavi launches World Studio, Stepfun debuts AI-native phone STEPS NEO; GPT-5.6 file deletion and AI companion shutdowns spark safety and regulation debates.

Jul 15, 2026·2 min

Mesh LLM: Pool Multiple Old Computers Into a Virtual Super GPU for Running Large Models

Mesh LLM is an open-source distributed inference framework that splits model layers across multiple devices, creating a virtual super GPU to run 100GB+ LLMs on consumer hardware.

JAX Host Offloading: A Practical Guide…

Jul 14, 2026·2 min

JAX Host Offloading: A Practical Guide to Breaking Through LLM Training Memory Bottlenecks

Memory capacity is the core bottleneck in LLM training. This guide explores JAX-based host offloading — covering optimizer state offloading, activation strategies, PCIe bandwidth trade-offs, and how it complements activation recomputation.

Jul 13, 2026·6 min

GPT-5.6 vs Grok 4.5 on the Same Day: A Packed Week of Flagship AI Model Launches

OpenAI's GPT-5.6 launches with Sawa, Terra, and Luna sub-models the same day as Musk's Grok 4.5, while Anthropic, Meta, and NVIDIA make their moves. A packed week of flagship AI launches.

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, 2026·6 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 12, 2026·5 min

GPT-5.6 Released: Outperforms Rivals but Halted by Government—Cybersecurity Capabilities Trigger Regulatory Red Line

OpenAI releases GPT-5.6 (Sol/Terra/Luna), beating Anthropic on Terminal Bench at ~40% lower cost. But its cybersecurity capabilities hit danger thresholds, limiting access to trusted partners at government request.

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, 2026·6 min

Why Is Qwen3 27B So Slow on a 4090? A Hands-On Guide to Local Inference Performance Optimization

RTX 4090 taking over 400 seconds to run Qwen3 27B inference? This article analyzes the core causes—VRAM overflow and CPU offloading drag—and offers targeted fixes.

Jul 10, 2026·5 min

Real-World Comparison of 6 Open-Source LLMs for On-Premise Deployment: A Selection Guide to Avoid Pitfalls

A hands-on comparison of 6 open-source LLMs (DeepSeek, Qwen3, Zhipu GLM, Kimi K2, MiniMax M3, Tencent Hunyuan 3) for on-premise deployment—covering hardware cost, inference efficiency, and deployment difficulty.

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

Unsloth Major Update: DeepSeek-V4 Support and a Deep Dive into NVFP4 Quantized Export

Unsloth v0.1.481-beta adds full DeepSeek-V4-Flash support, NVFP4/FP8/imatrix GGUF quantized export, 1.3x faster GRPO, 3-5x faster MoE training, and an OpenAI-compatible API service in Studio.

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, 2026·6 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.