ASUS Releases Servers with MI325X Accelerators Powered by AMD EPYC 9005 CPUs
A range of servers driven by the innovative AMD EPYC 9005-series CPUs were unveiled by ASUS today, establishing new benchmarks for performance and density for workloads in AI-driven data centers. Among the entire lineup is the AMD-compatible ASUS ESC A8A-E12U.
With the ability to accommodate eight GPUs for large-scale AI model training, the ASUS ESC8000A-E13P GPU servers, and Instinct MI325X accelerators provide unparalleled processing capability. One multi-node server for cloud computing and EDA is the ASUS RS520QA-E13. The RS720A-E13, RS700A-E13, RS521A, and RS501A are among ASUS’s adaptable solutions for general-purpose applications. To meet the expectations of the most demanding workloads, these servers are designed to give exceptional performance across a variety of applications.
At the heart of these servers are AMD EPYC 9005-series processors, formerly known as Turin. Based on the Zen 5 architecture, these powerful CPUs have up to 192 cores, 384 threads, and a frequency of up to 5GHz. They are meant to take AI-driven data center workloads to new heights. Superior performance in AI inference is achieved by EPYC 9005-series processors, which have an industry-leading vCPU density and combine enhanced computation and AI capabilities to handle both traditional and mixed jobs.
Because of the x86 architecture, which unifies this lineup diversity, clients can effortlessly incorporate ASUS servers with AMD EPYC 9005-series processors into their pre-existing x86 infrastructure. This makes it possible for the data center to migrate and transition in an efficient manner, guaranteeing continuity and increasing operational effectiveness.
AI servers: ESC A8A-E12U and ASUS ESC8000A-E13P
Specifically intended to provide parallel computing and flexible networking, the ASUS ESC A8A-E12U and ASUS ESC8000A-E13P are eight-GPU servers. With a 7U form factor, two AMD EPYC 9005-series CPUs, support for AMD Instinct MI325X accelerators, and a direct GPU-to-GPU connectivity, the ESC A8A-E12U enables effective scaling for big AI models and HPC workloads with up to 6TB/s of bandwidth.
Significantly improving support for NVIDIA OVE and NIM is made possible by the ASUS ESC8000A-E13P server’s strong AMD EPYC CPUs, generous memory, and complete adherence to the NVIDIA MGX architecture. Its characteristics allow customers to establish dedicated instances, speed up training and simulations, facilitate real-time collaboration, and guarantee that the infrastructure scales smoothly to accommodate the increasing demands of OVE and NVIDIA NIM projects, all of which promote additional innovation and adoption.
Scalable multi-node server: ASUS RS520QA-E13
The powerful ASUS RS520QA-E13 server is tailored for cloud computing and electronic design automation applications. Its cutting-edge AMD EPYC 9005-series processors are notable for their exceptional memory expandability, which allows for 12+8 DIMM combinations per node. 6000MTS for 1DPC and 4400MTS for 2DPC are made possible by this. Its creative front-access design provides versatility, and its ergonomic handle and tool-less bracket make maintenance simple. With its excellent thermal performance and effective air-cooling methods, the ASUS RS520QA-E13 is a great option for demanding workloads and can support processors up to 400W.
General-purpose servers: RS700A-E13, RS720A-E13, RS521A-E12 and RS501A-E12
With AMD EPYC 9005-series processors, ASUS RS720A-E13 and RS700A-E13 can support up to 24 DIMMs, have a maximum CPU TDP of 500W, and have up to 192 cores. For AI workloads and HPC, the GPU-optimized RS720A-E13 design supports up to 24 all-flash PCIe 5.0 NVMe drives and 3 dual-slot GPUs, guaranteeing effective processing of real-time data using trained models and precise forecasts. With up to 12 all-flash NVMe SSDs supported for increased bandwidth, the RS700A-E13 offers remarkable performance and efficiency for a variety of workloads.
Data centers with a variety of workloads can benefit from ASUS RS521A-E12 and RS501A-E12’s optimized power-usage effectiveness (PUE) and total cost of ownership (TCO). These servers are made to be scalable in order to meet the growing needs of contemporary data centers. With AMD EPYC 9005-series processors, they are ideal for dependable everyday business operations. They provide a variety of storage options and high memory capacity to improve acceleration for applications in on-premises and cloud data centers, as well as hybrid cloud environments.
A wide selection of ASUS devices using AMD power A variety of server solutions with 5th Generation AMD EPYC processors that can handle a variety of AI tasks are available from ASUS to accelerate generative AI and LLM applications. Whether it’s machine learning, financial computing, data search and analytics, or computational storage, the AMD platform guarantees amazing performance.
