AMD's AI Tech Premiere: Everything That Was Announced

Advanced Micro Devices, Inc. AMD on Tuesday announced artificial intelligence-focused advancements and offerings. The Data Center and AI Technology Premiere, led by CEO Lisa Su, introduced various innovations, including the AMD EPYC Bergamo processors and Instinct MI300 accelerators.

Here’s a look at everything AMD announced.

AMD EPYC Bergamo Chips

The AMD Bergamo boasts up to 128 cores per socket and 256 threads, marking a milestone for the company. Its increased core density will lead a new generation of hardware capable of supporting more demanding workloads.

The chips leverage the new Zen 4c efficiency cores, optimized for higher density while maintaining full feature set support.

Bergamo has high performance, nearing 1000 SPECrate2017_int_base per socket, meaning the processor is capable of handing a high number of integer calculations per second.

The chips cater to scale-out cloud instances rather than performance-per-core applications, making them versatile for various tasks. Bergamo chips are already shipping to AMD’s cloud customers.

Related: AMD CEO Lisa Su Announces Genoa As ‘The Best CPU For AI’ At Company’s Data Center Premiere

AMD Instinct MI300 Accelerators

AMD announced the Instinct MI300 accelerators, which have 146 billion transistors and are set to compete with Nvidia Corporation's NVDA Grace Hopper.

The MI300 blends 13 chiplets, with several 3D-stacked, creating a chip with 24 Zen 4 CPU cores fused with a CDNA 3 graphics engine and eight stacks of HBM3.

AMD Genoa-X Processors, Sienna Chips

AMD also announced the launch of the Genoa-X processors. The processors use a 3D-Stacked L3 cache to boost performance in technical workloads. Alongside Genoa-X, AMD provided updates on its first telecom-optimized chips, Siena, and news about the next-gen Zen 5 ‘Turin’ data center chips.

AMD’s Collaboration with Cloud Partners

AMD’s influence in the cloud space was also highlighted. Dave Brown, vice president of Amazon.com, Inc AMZN AWS’s EC2, mentioned the cost savings and performance advantages of using AMD’s instances in its cloud.

AWS is building new instances with AWS Nitro and the fourth-generation EPYC Genoa processors, offering 50% more performance than previous instances.

Oracle Corporation ORCL will also have Genoa E5 instances available in July.

Meta Platforms Inc META also confirmed plans to use Bergamo chips for its infrastructure.

The AI Market, AMD’s New Partnerships

Su emphasized the growing market opportunity for AI, driven by large language models, which could see the total addressable market reach approximately $150 billion by 2027, up from the current $30 billion.

AMD announced its partnership with New York-based private AI company Hugging Face, optimizing its models for AMD CPUs, GPUs, and other AI hardware.

The Future of AMD’s Data Center And AI

AMD’s approach to shaping the data center landscape involves further enhancing server manageability and reducing networking overhead.

The company developed the P4 DPU and SmartNICs, an integral part of the new data center architectures, in collaboration with Aruba Networks, whose parent company is Hewlett Packard Enterprise Co HPE.

Related: AMD And AWS Reveal Next-Gen Performance Leap In Cloud Computing At Tech Premiere

Photo courtesy of AMD.