AMD officially launched the Instinct MI300 AI accelerators for data centers and the ROCm 6 software package
December 10, 2023

AMD officially launched the Instinct MI300 AI accelerators for data centers and the ROCm 6 software package

Blog

At a specially organized “Advancing_AI” event, AMD hosted industry leaders including Microsoft, Meta, Oracle, Dell Technologies, HPE, Lenovo, Supermicro, Broadcom and Cisco to demonstrate how these companies are collaborating with AMD about on delivering advanced AI solutions that span from the cloud to the enterprise and the PC. AMD launched several new products at the “Advancing_AI” event, including the AMD Instinct MI300 series AI datacenter accelerator, ROCm™ 6 open source software suite with significant optimizations and new features supporting Large Language Models (LLM) and Ryzen™ 8040 processor series with Ryzen AI.

“AI is the future of computing and AMD is uniquely positioned to power the end-to-end infrastructure that will define this AI era, from massive cloud installations to enterprise clusters and intelligent AI embedded devices and PCs,” said AMD President and CEO Dr. Lisa Su. “We are seeing very strong demand for our new Instinct MI300 GPUs, which are the world’s highest performance accelerators for generative AI. We are also accelerating significant momentum for our data center AI solutions with the largest cloud companies, industry-leading server vendors and the most innovative AI startups working closely with us to quickly bring Instinct MI300 solutions to market that will dramatically accelerate the pace of innovation across the entire AI ecosystem .”

Instinct MI300 – Advancing AI data centers from the cloud to enterprise data centers and supercomputers

AMD was joined by multiple partners during the event to highlight the strong adoption and growing momentum for AMD Instinct data center AI accelerators.

· Microsoft has detailed how it implements AMD Instinct MI300X accelerators to run the new Azure ND MI300x v5 series of virtual machines (VMs) optimized for AI workloads.

· Meta noted that the company is adding AMD Instinct MI300X accelerators to its data centers in conjunction with ROCm 6 to power AI inference workloads and recognize the ROCm 6 optimizations that AMD has done on the Llama 2 model family.

AMD Instinct MI300 tekst 2AMD

· Oracle has unveiled plans to offer OCI bare-bones computing solutions featuring AMD Instinct MI300X accelerators, as well as plans to include AMD Instinct MI300X accelerators in its upcoming generative AI service.

· Major data center infrastructure providers have announced plans to integrate AMD Instinct MI300 accelerators into their product portfolios. Dell announced the integration of the AMD Instinct MI300X accelerator with their PowerEdge XE9680 server solution to deliver breakthrough performance for generative AI workloads in a modular and scalable format for clients. HPE announced plans to introduce AMD Instinct MI300 accelerators to its enterprise and HPC offerings. Lenovo has shared plans to bring AMD Instinct MI300X accelerators to the Lenovo ThinkSystem platform to deliver AI solutions across industries, including retail, manufacturing, financial services and healthcare. Supermicro has announced plans to offer the AMD Instinct MI300 GPU in its portfolio of AI solutions. Asus, Gigabyte, Ingrasy, Inventec, QCT, Wistron i Wiwynn they also plan to offer solutions based on AMD Instinct MI300 accelerators.

· Specialist AI cloud providers including Aligned, Arc Energy, Staircase, Crusoe, Denver Dataworks i Tensorwavesall plan to provide deals that will expand access to AMD Instinct MI300X GPUs for developers and AI startups.

Bringing an open, proven and ready AI software platform to market

AMD highlighted significant progress in expanding the software ecosystem that supports AMD Instinct data-center accelerators.

AMD has introduced the latest version of the open source software package for AMD Instinct GPUs, ROCm 6, which is optimized for generative artificial intelligence, especially for large language models. ROCm 6 boasts support for new data types, advanced graph and kernel optimizations, optimized libraries, and state-of-the-art attention algorithms, which together with the MI300X deliver a ~8x performance boost for total text generation latency on Llama 2 compared to ROCm 5 running on MI250.

AMD Instinct MI300 tekst 1AMD

· Databricks, Essential AI and Lamini, three AI startups building new models and AI solutions, joined AMD on stage to discuss how they are using AMD Instinct MI300X accelerators and the open ROCm 6 software suite to deliver differentiated AI solutions for business users.

· OpenAI adds support for AMD Instinct accelerators in Triton 3.0, providing out-of-the-box support for AMD accelerators that will allow developers to work at a higher level of abstraction on AMD hardware.

Continued AMD leadership in advancing AI on the PC

With millions of AI PCs shipped to date, AMD announced new flagship mobile processors with the launch of the latest AMD Ryzen 8040 series processors that deliver even more AI computing capabilities. AMD also launched Ryzen AI 1.0 – a software package that allows developers to easily implement applications that use pre-trained models to add AI capabilities to Windows applications.

AMD also revealed that the upcoming generation of “Strix Point” APUs, scheduled for launch in 2024, will include the XDNA™ 2 architecture that will deliver more than 3x AI computing performance gains over the previous generation enabling new generative AI experiences . Microsoft also joined to discuss how they are working closely with AMD on future AI experiences for Windows PCs.

Additional resources

  • View the complete AMD Advanced AI Keynote
  • Read more about the announcements during Advance AI here
  • Learn more about the AMD 30×25 Energy Efficiency Initiative here
  • Read here more information about the AMD Instinct MI300 series accelerators, ROCm 6 and the growing ecosystem of AMD AI solutions.