HomeWinBuzzer NewsMicrosoft Launches Cobalt Chips and AMD Accelerators for Azure

Microsoft Launches Cobalt Chips and AMD Accelerators for Azure

In a pre-Build 2024 meeting, Microsoft detailed the upcoming launch of the Cobalt 100 processor and its wider cloud computing strategy.

-

has announced the introduction of its custom Cobalt 100 chips to Azure, offering a public preview at the Build 2024 conference. The chips are expected to deliver a 40% performance improvement over existing ARM chips, according to Scott Guthrie, Microsoft's Executive Vice President of the and AI group. Microsoft introduced the Cobalt 100 last November

Comparison with AWS Graviton

In a pre-conference briefing reported by TechCrunch, Guthrie compared the new Cobalt chips to AWS's Graviton chips, which have been available for several years. The Cobalt chips, based on the Arm architecture, feature 128 cores and are designed to enhance performance significantly. Companies such as Adobe and Snowflake are already utilizing these chips in their operations.

Additional Hardware and Pricing Updates

Alongside the Cobalt chips, Microsoft will also introduce AMD's MI300X accelerators to Azure clients. Although AMD has historically lagged behind in the AI sector, the AMD MI300X is now considered a cost-effective alternative due to its enhanced software support. Guthrie described it as the most economical GPU for Azure OpenAI Service.

Microsoft will also lower the pricing for accessing and running large language models, though specific details will be revealed at the Build conference. Additionally, the company will preview a new real-time intelligence system for data streaming into Fabric, Microsoft's data analytics platform. This system will support native Kafka integration and compatibility with AWS Kinesis and Cloud's Pub/Sub.

Partnership with Snowflake and Copilot Integration

A new partnership with Snowflake will be also announced, allowing Fabric to support Snowflake's Iceberg format. This integration will ensure interoperability between Fabric and Snowflake, enabling data to be shared across both platforms.

For developers using , Microsoft plans to introduce a feature that allows management of Azure resources through natural language commands. This functionality aims to streamline the development process by enabling a tighter integration with Azure. The system is built on a common extensibility mechanism, allowing other providers to offer similar capabilities.

Microsoft's vice president of and artificial intelligence, Scott Guthrie, shared that his firm will introduce AMD's accelerated computing products to customers for its Azure cloud computing platform and announce custom variants of Microsoft's in-house Cobalt 100 chips at an upcoming conference.

At the analyst pre-conference, Guthrie highlighted Microsoft's Arm-based data center chips. Microsoft, like Google, Amazon, and parent Meta, has developed its custom processors to handle data center workloads. According to the Microsoft executive, when compared to other ARM-based products in the market, Microsoft's chip can offer as much as a 40% performance boost.

After Google's I/O this week, Microsoft's Build conference will enable it to showcase the latest products and technologies on its plate for artificial intelligence. The announcements will include the Cobalt processors, with Microsoft expected to highlight customized and previously unreleased variants of the 128 core chips.

Additionally, Guthrie added that AMD would also be a part of the Microsoft Build conference. Unlike NVIDIA, which faces constraints when it comes to designing processors, AMD can sell CPUs and GPUs. This allows it to provide a diversified product portfolio in the artificial intelligence market, and the firm's flagship product on this front is the MI300X accelerator.

The MI300X is a graphics card with the typical shading cores and other features that are specific to a GPU's design. It competes with NVIDIA's latest AI GPUs, namely the H100 products, and Microsoft will use the AMD GPUs to power up clusters for Azure customers.

Maia Processors and AI Workloads

Along with Cobalt, Microsoft also made another chip announcement in November 2023. Its other set of custom chips are the Maia processors. These are designed to specifically meet the needs of AI workloads and pack a considerable computing punch through 105 billion transistors.

While AMD does not sell pure AI processors, it offers support with its Ryzen processors. AMD's Ryzen AI software augments the CPUs and enables developers to run machine learning and other computing models on the chips.

Luke Jones
Luke Jones
Luke has been writing about all things tech for more than five years. He is following Microsoft closely to bring you the latest news about Windows, Office, Azure, Skype, HoloLens and all the rest of their products.