amd announced the products, strategy and ecosystem partners that claim to “shape the future of computing and highlight the next phase of data center innovation” at the company’s “Data Center and AI Technology Premiere.” The company was joined on stage by executives from Amazon Web Services (AWS), Citadel, Hugging Face, meta, Microsoft Azure And PyTorch to demonstrate the technology partnerships to bring the next generation of enhanced CPU and AI accelerator solutions to market.
Compute infrastructure optimized for data center
AMD has unveiled a series of updates to its 4th generation EPYC family. It is designed to provide customers with the workload specialization needed to meet the unique needs of businesses. The company also highlighted how the 4th Gen AMD EPYC processor will continue to boost performance and power efficiency. AMD was joined by AWS to preview next-generation Amazon Elastic Compute Cloud (Amazon EC2) M7a instances powered by 4th Gen AMD EPYC (“Genoa”) processors. outside the event, Oracle announced plans to make new Oracle Computing Infrastructure (OCI) E5 instances available with 4th generation AMD EPYC processors.
The company also introduced the 4th generation AMD EPYC 97X4 processors, previously codenamed “Bergamo”. With 128 “Zen 4c” cores per socket, these processors claim to provide the highest vCPU density, power efficiency, and performance for apps running in the cloud.
AMD was also joined by Meta, where the social media giant discussed how these processors are a good fit for their flagship apps like Instagram, WhatsApp, and more. Meta also saw performance improvements with 4th Gen AMD EPYC 97×4 processors compared to 3rd Gen AMD EPYC across different workloads. The latest processor offers TCO improvements, and the company explained how AMD and Meta optimized the EPYC CPUs for Meta’s power efficiency and compute density requirements.
AMD strives to make better products possible with technical computing
AMD introduced the 4th generation AMD EPYC processors with AMD 3D V-Cache technology, the world’s highest performing x86 server CPU for technical computing. Microsoft announced the general availability of Azure HBv4 and HX instances powered by 4th generation AMD EPYC processors with AMD 3D V-Cache technology.
AMD’s AI platform conveys its AI vision
The company also made a series of announcements to showcase its AI Platform strategy, giving customers a cloud advantage over its endpoint portfolio of hardware products. Deep industry software collaboration is needed to develop scalable and ubiquitous AI solutions.
AMD has also introduced the world’s most advanced generative AI accelerator. The company also revealed new details of the AMD Instinct MI300 Series accelerator family. This includes the introduction of the AMD Instinct MI300X accelerator, the world’s most advanced generative AI accelerator.
The MI300X is based on the next-gen AMD CDNA 3 accelerator architecture and supports up to 192 GB of HBM3 memory. This claims to provide the computing and memory efficiency needed for large language model training and inference for generative AI workloads.
With AMD Instinct MI300X’s large memory, customers can now fit large language models such as Falcon-40, a 40B parameter model, onto a single MI300X accelerator. AMD also introduced the AMD Instinct platformthat brings together eight MI300X accelerators in an industry-standard design for the ideal solution for AI inference and training.
AMD will test the MI300X with its key customers from the third quarter. The company also announced that the AMD Instinct MI300A, the world’s first APU accelerator for HPC and AI workloads, is now being sampled at customers.
The company also showcased the ROCm software ecosystem for data center accelerators. AMD also highlighted its willingness and collaboration with industry leaders to bring together an open AI software ecosystem.
PyTorch discussed work between AMD and itself to upstream the ROCm software stack and provide immediate “day zero” support for PyTorch 2.0 with ROCm release 5.4.2 on all AMD Instinct accelerators.
The integration proposes to provide developers with a wide range of AI models powered by
PyTorch that are compatible and ready to use “out of the box” on AMD accelerators.
Hugging Face, the open platform for AI builders, also announced that it will be optimizing
thousands of models on AMD platforms, from AMD Instinct accelerators to AMD Ryzen and AMD EPYC processors, AMD Radeon GPUs and Versal and Alveo adaptive processors.
A network portfolio for the cloud and enterprise
AMD also showed a network portfolio, including the AMD pensando DPU, AMD Ultra Low Latency NICs and AMD Adaptive NICs. In addition, AMD Pensando DPUs combine a software stack with “zero trust security” and an industry-leading programmable packet processor to create the world’s most intelligent and high-performance DPU.
The AMD Pensando DPU is deployed at scale with cloud partners such as IBM Cloud, Microsoft Azure and Oracle Compute Infrastructure. In the enterprise, it is deployed in the HPE Aruba CX 10000 Smart Switch and at customers such as leading IT services company DXC, and as part of VMware vSphere Distributed Services Engine, accelerating application performance for customers.
AMD even highlighted the next generation of its DPU roadmap, codenamed “Giglio”, which aims to provide customers with improved performance and energy efficiency compared to the current generation of products, when it is expected to be available in late 2023.
The company also announced the AMD Pensando Software-in-Silicon Developer Kit (SSDK), which enables customers to quickly develop or migrate services to deploy to the AMD Pensando P4 programmable DPU in coordination with the existing comprehensive feature set already deployed on the AMD Pensando P4 programmable DPU. AMD Pensando platform. The AMD Pensando SSDK enables customers to harness the power of AMD Pensando DPU leadership and customize network virtualization and security features within their infrastructure, in coordination with the existing set of features already deployed on the Pensando platform .
end of article