Author

Architects' Tech Alliance

Sharing project experiences, insights into cutting-edge architectures, focusing on cloud computing, microservices, big data, hyper-convergence, storage, data protection, artificial intelligence, industry practices and solutions.

2.3k

Articles

Likes

15.4k

Views

Comments

Latest from Architects' Tech Alliance

100 recent articles max

Architects' Tech Alliance

Sep 21, 2026 · Industry Insights

China's Domestic FC Switch Achieves 350ns Latency, Outpacing Broadcom's 580ns

Chengdu Guanghangxin's fully domestic 24-port 16Gbps FC SAN switch achieves sub-350ns latency, surpassing Broadcom's 580ns, leveraging lossless BB Credit flow control and crossbar architecture; the article details FC technology, market trends, and domestic breakthroughs in a market dominated by Brocade and Cisco.

BB CreditFC SANFibre Channel

0 likes · 18 min read

China's Domestic FC Switch Achieves 350ns Latency, Outpacing Broadcom's 580ns

Architects' Tech Alliance

Sep 19, 2026 · Industry Insights

2026 FMS: Flash Memory Industry Shifts to AI Memory Hierarchy with HBM, CXL, and HBF Standards

The 2026 Future of Memory and Storage conference reveals a decisive industry pivot from traditional SSD scaling to AI-centric memory architectures, highlighted by Samsung's zHBM stacking, Kioxia's GPU-direct XL-FLASH SSD, SK Hynix/WD's open HBF standard, and CXL-enabled memory pooling and tiering for next-generation AI data centers.

AI storageCXLFMS 2026

0 likes · 10 min read

2026 FMS: Flash Memory Industry Shifts to AI Memory Hierarchy with HBM, CXL, and HBF Standards

Architects' Tech Alliance

Sep 17, 2026 · Artificial Intelligence

Huawei Atlas 960E SuperPoD: First NPO Super Node for 10-Trillion-Parameter Models

Huawei unveils Atlas 960E SuperPoD, the industry's first super node using Near Package Optics (NPO) technology, scaling to 4,096 Ascend 960 chips delivering 8 EFLOPS FP8 and 16 EFLOPS FP4 for trillion-parameter model training, with 5,500 Hi-ONE optical engines replacing 48,000 modules to cut 550 kW power and achieve 99.8% availability.

AI super nodeAscend 960Atlas 960E

0 likes · 7 min read

Huawei Atlas 960E SuperPoD: First NPO Super Node for 10-Trillion-Parameter Models

Architects' Tech Alliance

Sep 17, 2026 · Artificial Intelligence

Moore Threads MTT C256 Supernode: 256-GPU Single-Layer Scale-Up Architecture & S5000 Chip Deep Dive

Moore Threads unveils the MTT C256 supernode at WAIC 2026, packing 256 GPUs into a single Scale-up network layer with compute-switch integration, breaking the 64-GPU limit. The MTT S5000 chip delivers 95,920 tok/s FP8 throughput, native FP8 KV Cache, and CUDA-source-level MUSA compatibility for heterogeneous prefill/decode inference.

FP8MTT C256MTT S5000

0 likes · 6 min read

Moore Threads MTT C256 Supernode: 256-GPU Single-Layer Scale-Up Architecture & S5000 Chip Deep Dive

Architects' Tech Alliance

Sep 16, 2026 · Industry Insights

Ningchang Unveils 1MW Liquid-Cooled Cabinet: 5000W Chip Cooling Redefines AI Infrastructure

Ningchang launches a 1MW single-phase cold-plate liquid-cooled cabinet achieving 5000W per-chip cooling and 1.2 LPM/kW flow efficiency, analyzing how MW-density infrastructure shifts AI data center economics from token-level optimization to power-thermal co-design.

AI infrastructureHigh-Density ComputingMW cabinet

0 likes · 7 min read

Ningchang Unveils 1MW Liquid-Cooled Cabinet: 5000W Chip Cooling Redefines AI Infrastructure

Architects' Tech Alliance

Sep 15, 2026 · Artificial Intelligence

Biren's 1024-GPU Super Node: NPO Optical Interconnect & Chiplet Architecture Explained

Biren Technology details a three-tier AI super node architecture scaling to 1024 GPUs via NPO optical interconnect and the BLink 2.0 protocol, built on its Chiplet-based BR20X GPUs with native FP8/FP4 support for trillion-parameter model training.

AI super nodeBLink 2.0BR20X GPU

0 likes · 6 min read

Biren's 1024-GPU Super Node: NPO Optical Interconnect & Chiplet Architecture Explained

Architects' Tech Alliance

Sep 14, 2026 · Artificial Intelligence

MetaX Xijing S600 Super Node: 64-GPU Cabinet, Zero-Cable OEX Architecture & MetaXLink Interconnect

MetaX unveils its Xijing S600 super node with 64 GPUs per cabinet, zero-cable OEX architecture, MetaXLink interconnect, and liquid cooling, alongside three other AI compute node form factors scaling from 16 to 128 GPUs for large model training and inference.

3D MeshGPU super nodeMetaX

0 likes · 8 min read

MetaX Xijing S600 Super Node: 64-GPU Cabinet, Zero-Cable OEX Architecture & MetaXLink Interconnect

Architects' Tech Alliance

Sep 13, 2026 · Artificial Intelligence

Alibaba Panjiu AL128: 128-Chip AI Super Node Achieves Pb/s Bandwidth, Sub-150ns Latency

Alibaba's Panjiu AL128 super node integrates 128 custom Zhenwu M890 AI chips in a liquid-cooled cabinet delivering Pb/s-scale bandwidth and sub-150ns P2P latency, using a decoupled open architecture with ICN Switch 1.0 and ALink System interconnect to boost inference performance 50% over traditional designs.

AI super nodeALink SystemAlibaba

0 likes · 8 min read

Alibaba Panjiu AL128: 128-Chip AI Super Node Achieves Pb/s Bandwidth, Sub-150ns Latency

Architects' Tech Alliance

Sep 12, 2026 · Artificial Intelligence

Ruijie RG-ETH 128 Supernode: Dual Architecture Scales AI Clusters to 1024 GPUs

Ruijie's RG-ETH 128 supernode uses a dual compute-switch architecture with 51.2 Tb/s switch chips to scale from 64 to 128 GPUs per cabinet via Host Tray decoupling, and up to 1024 GPUs across cabinets through 32×800G expansion ports, with a roadmap from 112G copper SerDes to 224G and LPO→NPO→CPO optical interconnects.

AI supernodeGPU clusterHost Tray

0 likes · 7 min read

Ruijie RG-ETH 128 Supernode: Dual Architecture Scales AI Clusters to 1024 GPUs

Architects' Tech Alliance

Sep 11, 2026 · Artificial Intelligence

scaleFabric: IBGDA-Powered GPU-Autonomous Communication Validated at 100K-GPU Scale

Sugon's scaleFabric achieves China's first large-scale IBGDA deployment, letting GPUs directly drive InfiniBand NICs without CPU involvement; validated in 100K-GPU clusters, it eliminates the CPU penalty for small-packet MoE All-to-All communication, cuts latency by two orders of magnitude, and integrates NVMe-over-RDMA and XPU Direct Storage for end-to-end Token acceleration.

All-to-AllGPUDirect RDMAIBGDA

0 likes · 11 min read

scaleFabric: IBGDA-Powered GPU-Autonomous Communication Validated at 100K-GPU Scale