AI on AMD in public cloud: Increased access at optimized cost & power (AIM230)

Here is a detailed summary of the video transcription in Markdown format:

Introduction to AMD

  • AMD is a ubiquitous technology company that provides support and technology for various applications like social media, streaming, search, video conferencing, collaboration, and productivity.
  • AMD is known for its high-performance data center CPUs, called EPYC, as well as its GPU product line, Instinct, for AI and HPC acceleration.
  • AMD also offers networking solutions like DPUs (Data Processing Units) and FPGAs (Field-Programmable Gate Arrays) for offloading and accelerating specific workloads.
  • AMD has seen a significant market share shift in the data center market, particularly with its 4th and 5th generation EPYC processors (codenamed Milan and Genoa), capturing over 34% market share in just 5 years.

Why Choose AMD in AWS

  • AWS launched its 7th generation instances based on AMD's 4th generation EPYC (Genoa) processors, which are designed to deliver the best x86 performance and price-performance in all of EC2.
  • The 4th generation EPYC (Genoa) processors offer a 50% performance uplift compared to the previous generation, enabling customers to run their applications with half the number of instances or instance sizes, leading to a 37% OPEX savings.
  • The performance and efficiency of AMD processors can significantly impact the overall OPEX, as infrastructure OPEX (general IT overhead) accounts for two-thirds of the total OPEX spend, and optimizing this can free up capital for investing in new technologies and business growth.

AI on AMD

  • AMD offers a range of products, including EPYC CPUs and Instinct GPUs, to address different AI workloads and use cases.
  • For training and large-scale real-time inference, AMD's Instinct GPUs are well-suited, while for mixed workload inference, small to medium models, and offline/batch inference, AMD's EPYC CPUs provide a practical and cost-effective solution.
  • AMD's EPYC CPUs offer leadership performance and efficiency in classical machine learning, recommendation systems, and language models, as demonstrated by various benchmarks and use cases.
  • AMD provides the Zen DNN plugin, which can be used with popular AI frameworks like PyTorch and TensorFlow, to further optimize the performance of AI workloads on AMD hardware.
  • The Zen DNN plugin leverages AMD's architecture-specific optimizations, including graph optimizations, multi-threading, and low-level kernel optimizations, to deliver up to 52% performance improvements across various AI workflows.

Closing Thoughts

  • AMD's EPYC CPUs can deliver 27-45% OPEX savings compared to competing solutions, enabling customers to reinvest those savings into innovation, business growth, or adopting new technologies like AI.
  • Customers are encouraged to visit the AMD booth at the conference to explore the cost and performance optimization benefits of AMD-powered instances and engage with AMD and partner experts.

Your Digital Journey deserves a great story.

Build one with us.

Cookies Icon

These cookies are used to collect information about how you interact with this website and allow us to remember you. We use this information to improve and customize your browsing experience, as well as for analytics.

If you decline, your information won’t be tracked when you visit this website. A single cookie will be used in your browser to remember your preference.

Talk to us