As demand for AI accelerator solutions rapidly grows, AMD, which is challenging Nvidia's stronghold in the AI market, held a meeting with key partners to explore future growth of AI solutions.

At its 'Advancing AI' event on the 7th, AMD announced various collaborations with major partners to provide advanced AI solutions to various sectors, including cloud, enterprise, and PC.

Microsoft, Meta, Oracle, Dell Technologies, HPE, Lenovo, Supermicro, Arista, Broadcom, and Cisco participated in this event.

AMD also announced a number of new products, including the Instinct MI300 series data center AI accelerator, the ROCm 6 open software stack with new features supporting large-scale optimization and large language models (LLM), and the Ryzen 8040 series processors with Ryzen AI.

Along with the launch of the latest AMD Ryzen 8040 series processors that provide AI computing capabilities, AMD also introduced 'Ryzen AI 1.0', a software stack that allows users to easily add and deploy AI capabilities to Windows applications with pre-trained models.

“AI is the future of computing, and AMD is uniquely positioned to power the end-to-end infrastructure that will define the AI era – from massive cloud to enterprise clusters, AI-enabled intelligent embedded devices and PCs,” said Lisa Su, CEO of AMD.

He also said, “We are seeing very strong demand for our Instinct MI300 GPU, the world’s highest-performance accelerator for generative AI,” adding that the pace of innovation across the AI ecosystem is accelerating, “We are building significant momentum in data center AI solutions with the largest cloud companies, the industry’s best server providers, and the most innovative AI startups.”

■ Expanding AI Ecosystem

During the event, AMD presented examples of adoption of AMD Instinct data center AI accelerators with several partners and highlighted its growth momentum.

Several large data center infrastructure providers have announced plans to integrate AMD Instinct MI300 accelerators across their product portfolios.
Microsoft has detailed how it will leverage AMD Instinct MI300X accelerators to power its new Azure ND MI300x v5 virtual machine series, optimized for AI workloads.

Meta said it is adding AMD Instinct MI300X accelerators to its data centers with ROCm 6 to power AI inference workloads, and is also aware that AMD has made ROCm 6 optimizations for its Rama 2 model family.

Oracle announced availability of OCI bare metal compute solutions featuring AMD Instinct MI300X accelerators and plans to include AMD Instinct MI300X accelerators in its future generative AI services.

Additionally, Dell is integrating server solutions powered by AMD products to deliver scalable, modular products that deliver breakthrough performance for generative AI workloads, and HPE announced plans to introduce the AMD Instinct MI300 accelerator.

Lenovo will introduce the AMD Instinct MI300X accelerator into its Lenovo ThinkSystem platforms, while Supermicro will offer the AMD Instinct MI300 GPU across its portfolio of AI solutions.

Launch of a Verified Open AI Software Platform

AMD also highlighted the expansion of the software ecosystem supporting its AMD Instinct data center accelerators.

AMD has unveiled the latest version of ROCm 6, an open-source software stack for AMD Instinct GPUs optimized for generative AI, particularly large-scale language models.

ROCm 6 is a new data type, supports advanced graph and kernel optimizations, optimized libraries, and powerful attention algorithms.

Comparing the overall latency of text generation based on Rama 2, ROCm6 running on MI300X provides up to 8x performance improvement over ROCm 5 running on MI250.

Startups developing the latest models and AI solutions, including Databricks, Essential AI, and Lamini, discussed how to provide differentiated AI solutions to customers by utilizing the AMD Instinct MI300X accelerator and the open ROCm 6 software stack.

OpenAI is adding support for AMD Instinct accelerators in Triton 3.0, enabling developers to perform higher-level abstraction tasks on AMD hardware.