#United States #Austin #Oracle Cloud #AI Performance #AMD Instinct

Oracle and AMD Collaborate to Revolutionize AI Performance on Cloud

In a groundbreaking announcement, Oracle and AMD have unveiled their collaboration aimed at transforming how businesses handle substantial AI workloads in the cloud environment. The integration of the AMD Instinct™ MI355X GPUs into Oracle Cloud Infrastructure (OCI) not only boosts performance but also provides customers with unprecedented options for large-scale AI training and inference.

The partnership is set to introduce zettascale AI clusters featuring an astonishing 131,072 MI355X GPUs. This robust setup is designed to facilitate extensive computing capabilities necessary for extensive AI model development, enabling users to build, train, and execute AI applications efficiently.

Key Benefits of the Collaboration

Significant Empowerment and Cost Efficiency

Mahesh Thiagarajan, EVP of Oracle Cloud Infrastructure, mentioned the commitment of both companies to enhance their customer experience. With this collaboration, customers can anticipate a performance boost exceeding 2.8X compared to earlier generations. This advancement not only makes AI deployments more efficient but also significantly reduces operational costs. The promise of improved price performance for AI workloads means that businesses can innovate without the hindrance of budget constraints.

Enhanced Hardware Specifications

The AMD Instinct MI355X GPU marks a substantial leap forward in computing technology, offering nearly triple the compute power of its predecessor, alongside a remarkable 50% increase in high-bandwidth memory. Each new GPU boasts 288 gigabytes of high-bandwidth memory 3 (HBM3), facilitating the execution of large models entirely in memory, thereby enhancing both inference and training speeds required for complex datasets.

Support for Cutting-Edge Applications

The integration of the latest floating point compute (FP4) standard allows OCI to support modern large language models and generative AI applications cost-effectively. This enables ultra-efficient, high-speed inference which is crucial in today’s fast-paced tech landscape.

Moreover, the introduction of a liquid-cooled design ensures maximum performance density to handle demanding AI workloads without compromising on data processing speeds. Customers can expect lower latency and faster training times even for the most complex applications.

Open-Source Flexibility

The partnership promotes an open-source architecture through the AMD ROCm platform, which allows users a seamless migration of existing code without the fear of vendor lock-in. This flexibility is essential for businesses aiming to leverage the robust capabilities of their cloud environments while utilizing established coding practices.

Advanced Networking Solutions

To top it off, Oracle will be the first to incorporate AMD Pollara AI NICs within their backend networks, which advances the overall network infrastructure. This integration not only enhances performance but also supports innovative functions such as programmable congestion control, crucial for high-throughput AI workloads.

Conclusion

The collaboration between Oracle and AMD signifies a pivotal evolution in cloud computing capabilities geared towards AI functionalities. By harnessing state-of-the-art technology, customers are now equipped to tackle their most challenging AI-related tasks with unprecedented efficiency and effectiveness. Businesses can confidently venture into AI operations, knowing they have the infrastructure to support their ambitious projects and innovations in the evolving landscape of artificial intelligence. As Oracle continues to pave the way for superior cloud offerings, the partnership with AMD exemplifies a commitment to scalable, high-performance computing solutions, setting a high standard in the industry.

For further information, customers and interested parties can delve deeper into the advancements of OCI and the power of AMD Instinct MI355X GPUs on their respective websites.