#amdepyc

Stop falling for hardware marketing illusions.

Every time chip manufacturers drop new silicon, marketing teams try to sell it as a universal magic bullet. The new AMD EPYC 8005 (Sorano) is officially pitched for "edge computing," but hosting providers are blindly trying to sell it as the ultimate dense private cloud platform.

That is a catastrophic engineering lie.

Here is the actual physical truth about deploying this hardware:

The Virtualization Death Trap: You get 84 physical cores, but AMD intentionally limited this chip to exactly 6 memory channels. If you pack this thing with heavy, independent virtual machines, 14 cores end up violently fighting for a single memory lane. The bandwidth starves. Your cluster chokes.

The Container & Storage Masterpiece: Where does it actually win? Lightweight Linux containers (Kubernetes/Docker) and Software-Defined Storage. Containers share the host kernel, completely bypassing that memory bandwidth starvation. Plus, with 96 PCIe Gen 5 lanes, it’s an absolute monster for massive NVMe arrays (Ceph/MinIO).

The Base Clock Physics: To fit 84 cores into a strict 225W power envelope, the base clock is locked at 1.6 GHz. Do not run single-thread dependent applications (like game servers) on this. It is built strictly for massive, highly parallel background workloads.

At iRexta, we don't misallocate hardware. We use the EPYC 8005 exactly for what it was built for: scalable Kubernetes environments and high-throughput bare metal storage clusters.

Stop paying public cloud premiums for virtualization bottlenecks.

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.

AMD EPYC 9005 Processors are the driving force of MSI DC MHS Servers

Today, MSI unveiled its newest AMD EPYC 9005 Series CPU-based server boards and platforms, designed to meet the most demanding workloads in data centers with exceptional efficiency and performance.

MSI’s new server platforms, which feature AMD EPYC 9005 Series processors with up to 192 cores and 384 threads, offer revolutionary compute power, unmatched density, and remarkable energy efficiency. These features make them perfect for managing workloads that are cloud-native, AI-enabled, and business-critical in contemporary data centers.

MSI’s Enterprise Platform Solutions General Manager Danny Hsu stated the AMD EPYC 9005 Series CPU-based solutions offer a key advance for data centers seeking operational efficiency, and minimize their environmental footprint. These solutions guarantee that businesses can promptly adjust to new possibilities and remain ahead of changing market needs.

What is DC-MHS?

The Open Compute Project (OCP) created DC-MHS to produce modular, scalable datacenter hardware. It supports hyperscale, corporate, and edge datacenters by allowing hardware component design and integration flexibility.

In big datacenters, DC-MHS standardizes and streamlines hardware interfaces to simplify upgrades and maintenance. The program contains the DC-Stack and DC-SCM standards and addresses thermal control, power efficiency, and error recovery. These meet rising datacenter hardware performance and efficiency needs.

DC-MHS Server Solutions Redefining Modern Data Center Flexibility and Scalability

The DC MHS server systems from MSI meet the changing demands of modern data centers. With the help of the DC-SCM2 server management module, which features the Aspeed AST2600 BMC, and AMD EPYC 9005 Series processors, the DC-MHS series greatly improves server maintainability and operational efficiency, allowing businesses to deploy high-performance systems with the flexibility required to remain competitive in the quickly evolving market of today.

DC-MHS Multi-Node Servers

Modern data centers may improve operational efficiency and future-proof their infrastructure with MSI’s DC MHS multi-node server systems, which house several nodes inside a single chassis while optimizing space and resource use and lowering operating costs. A single AMD EPYC 9005 Series CPU with a maximum TDP of 500W powers each platform. It also has 12 DDR5 DIMM slots for high memory bandwidth and an OCP NIC for improved connectivity.

With support for 12 PCIe 5.0 x4 U.2 NVMe bays per node, the CD370-S4051-X2 3U 2-Node Core Compute Server offers quick data transmission and high-speed I/O, making it perfect for workloads involving a lot of data and AI.

With six PCIe 5.0 x4 U.2 NVMe bays per node, the CD270-S4051-X2 2U 2-Node Core computing Server offers quick, scalable computing performance that is perfect for virtualization and data analytics.

Features of the CD270-S4051-X4 2U 4-Node Core Compute Server Nodes with three PCIe 5.0 x4 U.2 NVMe bays each and liquid-cooled alternatives perform very well in applications requiring high throughput, including cloud settings.

DC-MHS Enterprise Server

2U Enterprise Server CX271-S4056: Fast, scalable performance is provided by a single AMD EPYC 9005 Series CPU, 24 DDR5 DIMM slots with a maximum speed of 5200MT/s, 8 or 24 2.5-inch PCIe 5.0 x4 U.2 NVMe bays, and PCIe 5.0 x16 OCP NIC ports. In contemporary data centers, its adaptable architecture which includes GPU acceleration options makes it perfect for data-intensive workloads like artificial intelligence, virtualization, and cloud-native apps.

DC-MHS Server Boards

D4056 DC MHS M-DNO Type-4 HPM: Designed for memory-intensive and data-driven applications, this platform offers quick data access and supports a single AMD EPYC 9005 CPU, 24 DDR5 DIMM slots, 2 NVMe M.2 ports, PCIe 5.0 extension slots, and OCP 3.0 NIC support.

Type-2 HPM D4051 DC MHS M-DNO: This board offers scalable performance for latency-sensitive applications in a half-width form factor and is designed for power economy. It has a single AMD EPYC 9005 CPU, 12 DDR5 DIMM slots, 2 NVMe M.2 ports, and a PCIe 5.0 x16 OCP NIC slot.

CXL 8TB Memory Expansion Server for Data-Intensive Workloads

Memory Expansion Server S2301 2U CXL: The server, which has 24 DDR5 DIMM slots and two AMD EPYC 9005 Series CPUs, is designed to satisfy data centers’ needs for unmatched scalability and memory capacity. Utilizing cutting-edge CXL 2.0 technology, the S2301 incorporates eight E3.S 2T drive bays for CXL memory and eight E3.S 1T NVMe storage, increasing the total memory capacity to an astounding 8TB. This makes it the perfect choice for data-intensive applications like High-Performance Computing (HPC), Electronic Design Automation (EDA), and In-Memory Database.

4GPU GPU servers for high-performance computing and artificial intelligence

The GPU servers G4101-01 and G4101-03 provide outstanding AI, machine learning, and HPC performance. These servers include four PCIe 5.0 triple-wide slots that enable GPGPUs up to 450W and twelve DDR5 DIMM slots. They are powered by a single AMD EPYC 9005 Series CPU with 500W TDP and liquid cooling. They provide the dependability and power needed for the most taxing applications with to their 12 PCIe 4.0 U.2 NVMe SSDs and Platinum or Titanium redundant 3000W power supply.

Enterprise Performance-Critical Workloads and Smaller Data Center Storage Servers

With up to 24 DDR5 DIMM slots and outstanding scalability, speed, and memory bandwidth of up to 6000MT/s, MSI’s Enterprise and Storage Server portfolio now supports AMD EPYC 9005 Series CPUs.

High-speed data transmission is designed for the S2206-05-10G, S2206-06-10G, and S1206-02-10G enterprise servers. These models are designed for tasks that need high performance, including AI and machine learning, and come with 10G SFP+ Ethernet. The top-performing model in this series, the S2206-06-10G, is designed for data-intensive applications that need the greatest possible I/O throughput. It offers remarkable storage capacity and performance with its three PCIe 5.0 x16 FHHL slots, sixteen PCIe 4.0 U.2 NVMe SSDs, and eight SATA/SAS bays. Storage servers like the S2206-04 provide affordable solutions with robust compute and storage performance at a reasonable price point, making them perfect for smaller data centers in settings where 1 Gbps network bandwidth is enough.

AMD Presents the Embedded 8004 Series, an Energy-Efficient EPYC for Systems. AMD EPYC Embedded CPUs regularly set the industry standard for networking, storage, and industrial applications due to their performance, efficiency, connection, and innovation. It maintain this leadership with the fourth-generation AMD EPYC Embedded 8004 Series CPUs.

AMD EPYC Embedded 8004 Series

AMD EPYC Embedded 8004 Series processors are suited for compute-intensive embedded systems with limited space and power due to their tiny form factor and strong performance. Additionally, it has several integrated functions to improve system efficiency and reliability. The AMD EPYC Embedded 8004 Series excels in the most demanding embedded environments. Corporate and cloud warm/cold storage, networking systems, routers, security appliances, and industrial edge applications benefit from it. Dynamic workloads are also easy for it.

Breakthrough performance and power efficiency

AMD EPYC Embedded 8004 Series processors boost performance-per-watt and core density using AMD “Zen 4c” cores. The EPYC Embedded 8004, AMD’s first processor series with these cores, sets a new bar for platform innovation and efficiency. Hardware manufacturers may now create unique, energy-efficient platforms with up to 30% more performance per watt than the previous generation (“Zen 3”) with this breakthrough.

These processors are intended to fulfill a variety of application demands. They are available in 1P configurations with 12 to 64 cores (24 to 128 threads) and support up to 1.152TB DDR5 memory capacity (2 DIMMs/channel with 96GB DIMM size). Their Thermal Design Power (TDP) profiles range from 70W to 225W.

Sizeable I/O And Function Packed In A Small Footprint

Because of its extensive memory capacity (6 channels of DDR5-4800) and high-speed I/O connectivity (96 lanes PCIe Gen 5), AMD EPYC Embedded 8004 Series processors are designed to easily handle data-intensive tasks. System designers may easily link SSDs, networking cards, and other components with these characteristics to construct scalable and adaptable system setups.

They are more energy-efficient and take up less room because to their tiny, SP6 socket form factor, which is 19% smaller than AMD EPYC Embedded 9004 Series processors. In addition, the devices come with a seven-year lifetime warranty, which aids system designers in preserving platform lifespan.

Improved Embedded Functions

The AMD EPYC Embedded 8004 Series CPUs are exceptional in terms of data preservation, system dependability, and data transport capabilities. Important characteristics that are incorporated include:

Direct Memory Access (AMD EPYC 4th Gen DMA): Intended to increase system speed and efficiency by relieving the CPU of data transfers, freeing up cores for important application activities.

Non-Transparent Bridging (NTB): By facilitating data sharing between two CPUs in active-active configurations via PCI Express (PCIe), Non-Transparent Bridging (NTB) improves system dependability and allows for sustained operation in the event of a breakdown.

DRAM Flush to NVMe: In the case of a power outage, this technique helps to guarantee that important data is kept safe by flushing it from DRAM to nonvolatile memory.

Dual SPI Support: Offers an extra degree of protection by allowing the use of two SPI ROMs, one for the secure bootloader and one for the BIOS image.

Device Identity Attestation: Enables cryptographic processor authentication, which helps prevent illegal CPU upgrades.

Support for the Yocto Framework: Enables users to design a thin, efficient Linux operating system for embedded devices.

Processors from the AMD EPYC Embedded 8004 Series are perfect for markets where high performance is required while maintaining strict standards for platform compactness, thermal dexterity, and energy economy. AMD EPYC Embedded 8004 Series processors are designed to meet the demands of demanding workloads in harsh environments, especially in networking, storage, and industrial edge systems. These Processors provide customers cutting-edge capabilities and unprecedented performance-per-watt advantages.

Uses

The Optimal Blend: Highest Efficiency and Maximum Performance.

Security and Networking

AMD Embedded networking solutions provide a comprehensive, unified architecture across the whole security product range by extending the performance and scalability of the client platforms.

Keepsake

High-performance storage processors with a wealth of capabilities are provided by AMD Embedded storage solutions, making storage solutions dependable and effective for embedded applications.

Commercial

For industrial control applications, AMD Embedded Industrial Control solutions provide embedded processor solutions with outstanding performance, a wide range of capabilities, and consistent long-life availability.

Summary

The AMD EPYC Embedded 8004 Series processors deliver unprecedented performance-per-watt with extensive I/O agility by using the performance and efficiency advantages of the “Zen 4” and “Zen 4c” core architecture (5nm).

These CPUs have outstanding compute performance and comprehensive I/O capabilities, making them perfect for next-generation networking, security, storage, and industrial applications. They are optimized for demanding workloads and “always on” embedded systems.