AMD Expands AI Cloud Infrastructure Partnership with Oracle; Sixth-Generation EPYC Processors to be Used in Compute E7 Cloud Instances

·by Henderson·AI
AMD Expands AI Cloud Infrastructure Partnership with Oracle; Sixth-Generation EPYC Processors to be Used in Compute E7 Cloud Instances

AMD has announced an expansion of its collaboration with Oracle in artificial intelligence and cloud infrastructure. According to an official announcement from AMD, the upcoming Compute E7 cloud computing instances from Oracle Cloud Infrastructure (OCI) will utilize AMD's sixth-generation EPYC server processors, while AMD's Pensando data processing units (DPU) will provide the technical foundation for Oracle's Acceleron converged network cards, serving OCI's computing instances equipped with AMD processors and accelerators.

This collaboration between AMD and Oracle involves both CPU computing and network infrastructure. The Compute E7 will utilize AMD's sixth-generation EPYC processors, aimed at supporting enterprise applications and AI model execution tasks, including agent scheduling, parallel operation of multiple agent sandboxes, information retrieval, memory caching, web services, and API services. As AI applications shift from merely generating content to invoking tools, querying databases, and executing tasks, the peripheral workload on CPUs is also increasing.

AMD and Oracle's Collaboration Will Enhance Cloud Computing Performance

According to Oracle's official technical introduction, the sixth-generation AMD EPYC processors utilize an architecture codenamed Venice, with each processor offering up to 256 cores and 512 threads, equipped with up to 1024MB of level 3 cache, and supporting 16 memory channels and PCIe 6.0. The Compute E7 will also provide flexible CPU and memory configurations, allowing enterprises to adjust resource allocation based on specific application computing needs.

On the networking front, AMD's Pensando DPU will be used in Oracle's Acceleron converged network cards. This solution allows network, storage, and security infrastructure tasks to be handled by dedicated hardware, reducing the demand on the host CPU's computing resources, enabling the CPU to allocate more capacity to enterprise applications and AI tasks. AMD's expanded collaboration with Oracle also extends their previous GPU computing partnership into a more comprehensive infrastructure system.

Oracle has already provided cloud computing instances based on AMD Instinct MI300X and MI355X, while AMD Helios further integrates GPUs, memory, high-speed interconnects, and ROCm software to provide a rack-scale computing platform for large-scale AI training, inference, and fine-tuning.

ItemSpecification
ProcessorAMD Sixth-Generation EPYC
Core CountUp to 256 cores
Thread CountUp to 512 threads
Level 3 CacheUp to 1024MB
Memory Channels16 channels

H
About the author
Henderson