Skip to main content

ACCWCards Comparison Roadmap

Chip Architectures

GPU (General Parallel)
TPU (Tensor Processing)
LPU (Language Processing)
NPU (Neural Processing)
APU (Accelerated Processing)
ASIC (Application-Specific)
WSE (Wafer-Scale)
IPU (Intelligence Processing)
RPU / RDU (Reconfigurable Dataflow)
PIM (Processing-in-Memory)
Neuromorphic (Spiking Neural)

All Posts
Product Launches
Vendor Strategy
Tech Deep Dives
Selection Guides
Industry News
Archive

中文
English

Post Categories

2026

AMD Advancing AI 2026前瞻：MI450已送样、Helios机架下半年放量，OpenAI+Meta锁定12GW大单
国产GPU上市潮：四小龙资本市场集结，摩尔线程MTT S5000对标H100
WAIC 2026前瞻：华为Atlas 950 SuperPoD真机首展，国产算力"镇馆之宝"集结
华为昇腾950系列产能与订单深度：950PR月产能跳升10倍，字节56亿美元锁定35万片
NVIDIA Vera Rubin正式出货：首台VR200 NVL72交付，三星HBM4量产，Rubin Ultra机柜天价
2026年H1 AI芯片行业复盘：Blackwell Ultra、国产三强与推理新时代
2026年6月AI芯片重大事件汇总：昇腾910C训练万亿模型、OpenAI自研芯片、RTX Spark发布
OpenAI 自研 AI 芯片 Jalapeño 深度解析：9 个月流片，推理成本降低 50%
国产 AI 芯片三巨头对比（2026）：昇腾、寒武纪、摩尔线程，谁是中国版 H100？
寒武纪 MLU690 vs NVIDIA H100 深度对比：国产 AI 芯片能否替代 H100？
2026 H2 AI 芯片路线图重大更新：Qualcomm 入局、AMD MI400 三款型号揭晓、华为三代路线图
AMD MI455X CES 2026 震撼发布：4年AI芯片性能涨1000倍
谷歌TPU 8i/8t 正式发布：训练与推理首次分家，2nm工艺赋能智能体时代
里程碑！华为昇腾910C完成1.6万亿参数大模型全参数训练
Intel Gaudi 4 / Jaguar Shores 最新进展：重返AI竞赛，HBM4内存加持
NVIDIA Vera Rubin 全面投产：智能体AI工厂时代正式开启
2026年国产AI芯片新进展：华为昇腾950、昆仑芯M100、阿里平头哥M890全面解析
华为昇腾910C深度解析：规格、部署与性能全览
昆仑芯P800深度解析：性能数据、架构创新与超节点部署
AI硬件进入"落地纪元"丨2026年五大变革与生存法则
Computex 2026 Wrap-Up: AI PC Chip War Begins, NVIDIA RTX Spark Arrives Fall 2026
Computex 2026 AI Compute Card Major Events: DGX Station for Windows, Intel Crescent Island, and More Major Launches
Huawei Ascend 950 Mass Production and the Full Picture of China's AI Chip Ecosystem
2026 H2 Top AI Chip Selection Guide: From H100 to Rubin, MI400, TPU 8t, TPU 8i
2026全球AI算力报告及算力产业十大趋势重磅发布
NVIDIA Launches RTX Spark: AI Compute Enters the Personal Computer Era
AI Cluster Power Crisis: 1MW Racks, Nuclear Plants, SMRs, and Green AI
AI Chip Startup Survival Report: Tenstorrent / SambaNova / Graphcore in 2026
HBM Three-Way Battle: SK Hynix / Samsung / Micron Fight for AI Memory Supremacy
Rack-Scale AI Era: NVL72 vs Helios vs Groq 3 LPX vs Trn3 UltraServer — Four Major Solutions Compared
Intel Cancels Falcon Shores, Pivots to Jaguar Shores: From Single-Chip Competition to Rack-Scale Systems
Inference Optimization Technology Evolution: PagedAttention / FlashAttention / Speculative Decoding Deep Dive
Apple Silicon Comeback: M3 Ultra 192GB UMA Local LLM Revolution
AMD MI400 + Helios Rack: 432GB HBM4 + 260 TB/s UALoF Open Interconnect
Google TPU 8t + 8i: The First TPU Generation to Split Training and Inference
NVIDIA Vera Rubin Platform Deep Dive: 6-Chip Package, 288GB HBM4, 50 PFLOPS FP4
Cerebras IPO Deep Dive: S-1 Filing, $22-25B Valuation, OpenAI $10B Mega-Deal
NVIDIA Acquires Groq for $20 Billion: LPU Officially Enters the NVIDIA Ecosystem

2025

AWS Trainium 3 GA: 3nm Process + 4.4× Compute + 4× Efficiency + 144-Chip UltraServer
Huawei Ascend 920: China's Highest Bandwidth at 4 Tbps + 3× H20 Compute for Domestic Substitution
China AI Chip Landscape 2025: Ascend, Cambricon, Hygon — Who Will Dominate?
GPU vs NPU vs TPU: In-Depth Comparison of Three AI Accelerator Architectures — Which One Should You Use?
AI Accelerator Selection Guide 2025: From Training to Inference — How to Choose the Best Chip?
MirrorFrog Update: New Blog and Glossary, Major Content Expansion
MirrorFrog Site Launch — AI Accelerator Driver and Documentation Directory

Authors

AI Compute Cards Wiki Editorial
40
Industry Research Team
Charles Qing
5
AI Hardware Analyst

By Type

AI Training GPU
AI Training ASIC
Wafer-Scale Training
AI Inference GPU
AI Inference ASIC
Workstation / Consumer
Edge AI
Innovative

By Vendor

NVIDIA
AMD
Intel (Habana)
Google TPU
AWS
华为昇腾
Cerebras / 其他

By Architecture

GPU / TPU / LPU
NPU / APU / ASIC
WSE / IPU / RPU
PIM (Processing-in-Memory)
Neuromorphic (Spiking)

Resources

Full Comparison
Future Roadmap
Blog

Copyright © 2026 AI Compute Cards Wiki. Built with Docusaurus.