AI Chronicle AI编年史 [秋芝2046]
ChatGPT诞生后的700多天，世界发生了什么?
More than 700 days after the birth of ChatGPT, what happened to the world?
AI很单纯，复杂的是人～
我是秋芝，欢迎来到2046～
Colossus is an AI training cluster developed by Elon Musk's artificial intelligence company, xAI, in Memphis, Tennessee:
The cost of Colossus Phase 1 is estimated to be between $3-4 billion. The cost for an Nvidia H100 is approximately $25,000, and buying 100,000 is going to set you back $2.5 billion (unless Nvidia is giving volume discounts).
Features
Colossus is one of the world's most powerful AI training platforms, with 100,000 Nvidia H100 GPUs. It's designed to train xAI's large language model, Grok, and compete with industry leaders like OpenAI and Google.
Building
Colossus was built in 122 days and uses a variety of components, including:
Racks: The racks are made by Supermicro and Dell, and are mostly liquid cooled. Each rack has eight nodes, with 64 GPUs per node.
Storage: Storage is provided by VAST Data on Supermicro and possibly Dell hardware.
Networking: The GPUs are linked using the Nvidia Spectrum-X ethernet networking platform.
Cooling
The racks use rear-door heat exchangers to capture heat that isn't covered by the cold-plate cooling.
Impact
The Colossus project has been the subject of controversy, with some campaigners criticizing its impact on air quality and the amount of water used to cool the data center.
No comments:
Post a Comment