Sign up to save tools and stay up to date with the latest in AI
bg
bg
1

Google’s new Trillium AI chip delivers 4X speed and powers Gemini 2.0

Dec 11, 2024 - venturebeat.com
Google has introduced Trillium, its sixth-generation AI accelerator chip, which significantly enhances AI development by delivering four times the training performance of its predecessor while using less energy. This chip powered the training of Google's Gemini 2.0 AI model and is part of a massive deployment of over 100,000 chips, creating one of the world's most powerful AI supercomputers. Trillium offers a 4.7x increase in peak compute performance, doubles memory capacity and bandwidth, and improves energy efficiency by 67%. These advancements are expected to transform AI development economics, offering up to 2.5x improvement in training performance per dollar, which could benefit enterprises and startups alike.

The release of Trillium marks a strategic move by Google in the competitive AI hardware market, challenging Nvidia's dominance with its custom silicon approach. By making Trillium available to cloud customers, Google aims to compete more aggressively in the cloud AI market against Microsoft Azure and Amazon Web Services. The chip's ability to efficiently handle mixed workloads suggests a future where AI computing becomes more accessible and cost-effective. This development signals a new phase in the race for AI hardware supremacy, with Google investing heavily in infrastructure to support the next generation of AI advancements.

Key takeaways:

```html
  • Google has unveiled Trillium, its sixth-generation AI accelerator chip, which offers significant performance improvements, including a 4x increase in training performance over its predecessor.
  • Trillium's enhanced energy efficiency and cost-effectiveness could reshape the economics of AI development, providing up to 2.5x improvement in training performance per dollar.
  • Google has deployed over 100,000 Trillium chips in a single network fabric, creating one of the world's most powerful AI supercomputers, highlighting its integrated approach to AI infrastructure.
  • The release of Trillium intensifies competition in AI hardware, challenging Nvidia's dominance and signaling a new phase in the race for AI hardware supremacy.
```
View Full Article

Comments (0)

Be the first to comment!