联想发布“AI工厂”解决方案 升级联想万全异构智算平台4.0

Core Insights - Lenovo Group officially launched the "Lenovo AI Factory" solution and introduced the next-generation high-end AI server, Lenovo Wenti WA8080a G5, based on Intel® Xeon® 6 processors, along with the upgraded Lenovo Wanquan Heterogeneous Computing Platform 4.0 [1][7] Group 1: AI Factory Solution - The "AI Factory" solution aims to create a standardized system that is manageable, replicable, and supports continuous operation, transforming complex and isolated AI development tasks into a modernized "AI production line" [5] - The solution processes customer data as raw materials through an intelligent agent development platform and AI training engine, ultimately producing deliverable intelligent agents, domain models, and inference services [5] Group 2: Technological Upgrades - The Lenovo Wanquan Heterogeneous Computing Platform 4.0 features four major technological upgrades: 1. In the pre-training scenario for large models, training time can be reduced by 35% through long-sequence parallel optimization [6] 2. In the post-training scenario, optimizations can shorten training time by 50% [6] 3. In the inference scenario, bandwidth utilization can be improved by 60%, and inference performance can be enhanced by 30% [6] 4. In the super-intelligent fusion computing scenario, it supports both international and domestic hardware ecosystems, optimizing from the underlying architecture to operators [6] Group 3: Collaboration and Performance - Lenovo's Wanquan Heterogeneous Computing Platform 4.0 has provided comprehensive support for various computing scenarios, including collaboration with the national-level high-quality AI cluster and improving model floating-point operation utilization from 30% to 60% [6] - The DeepSeek R1 model's throughput has exceeded 12,000 tokens per second in enterprise AI infrastructure scenarios [6]