Amazon announces new chips for training and running AI models

Amazon has announced the availability of two new chips designed for training and running AI models. The chips, called Trainium2 and Inferentia2, are the latest additions to Amazon's growing portfolio of custom silicon.

Trainium2 is designed to be up to 4x faster and 2x more energy efficient than the previous generation Trainium chip. It is also the first Amazon Web Services (AWS) chip to support the BFloat16 data format, which can help to reduce the training time of large language models by up to 3x.

Inferentia2 is designed to be up to 2x faster than the previous generation Inferentia chip. It also supports a wider range of data formats, including INT8, FP16, and BFloat16. This makes it a more versatile chip for running a variety of AI models.

The new chips are available now in Amazon Elastic Compute Cloud (EC2) instances. They are also available in Amazon SageMaker, AWS's machine learning platform.

Here are some of the benefits of using Trainium2 and Inferentia2:

• Increased training speed: Trainium2 can help to train large language models up to 3x faster than the previous generation Trainium chip. This can save businesses time and money when developing new AI models.
• Reduced training costs: Trainium2 is 2x more energy efficient than the previous generation Trainium chip. This can help businesses to reduce their cloud computing costs.
• Improved inference performance: Inferentia2 is up to 2x faster than the previous generation Inferentia chip. This can help businesses to run their AI models in real time.
• Wider range of data formats: Inferentia2 supports a wider range of data formats than the previous generation Inferentia chip. This makes it a more versatile chip for running a variety of AI models.

Overall, Trainium2 and Inferentia2 are powerful new chips that can help businesses to accelerate their AI development and deployment.

In addition to the new chips, Amazon also announced a number of other updates to its AI services, including:

• Amazon SageMaker Canvas: A new tool that makes it easier for businesses to develop and deploy AI models without writing code.
• Amazon SageMaker Autopilot: A new tool that automatically selects the best algorithms and hyperparameters for a given dataset.
• Amazon SageMaker Edge: A new service that makes it easier to deploy AI models to edge devices.

These updates are designed to make it easier for businesses of all sizes to take advantage of AI.