Engineering richer features in a revolutionary platform, AMD has made an incredible jump in processor performance in the business transformation journey. AI and deep learning performance will benefit from the built- in AI acceleration engines, while networking, storage, and analytics leverage other specialized accelerators in the 4th & 5th Gen AMD EPYC processors.
Engineering richer features in a revolutionary platform, AMD has made an incredible jump in processor performance in the business transformation journey. AI and deep learning performance will benefit from the built-in AI acceleration engines, while networking, storage, and analytics leverage other specialized accelerators in the 4th & 5th Gen AMD EPYC processors. Adding a host of new features to target a wide range of workloads, the new families of AMD EPYC processors will deliver even better CPU performance and performance per watt, and do so on a PCIe 5.0 platform with 2x the prior gen throughput to greatly speed up data movement to and from GPUs and storage. Intel also created the AMD EPYC CPU 9004/9005 Series with High Bandwidth Memory (HBM) for improvements in memory bound HPC and AI workloads. For this new platform, HEXADATA has products ready to get the most out of AMD EPYC CPU-based systems that support fast PCIe Gen5 accelerators and Gen5 NVMe drives, in addition to support for high performant DDR5 memory.
The SP5 socket has reached its pinnacle form with 5th Generation AMD EPYC™ processors. The same successful platform that was shared by AMD EPYC™ 9004 & 8004 series processors is now the same one for AMD EPYC™ 9005 series processors. With up to 192 cores, increased frequencies and cache, this top tier performance platform targets general purpose, cloud native, and technical computing. Building on the advantages of the EPYC™ 9004 series, the new EPYC™ 9005 series adopts the 3nm process with AMD "Zen 5" and "Zen 5c" core architecture, excelling in both energy efficiency and cost optimization. Hexadata has already prepared for this new processor with new servers and updates to existing products for 4th Gen AMD EPYC™ processors.
A PLX switch is required for multi-GPU systems, particularly in high-performance computing (HPC) servers and workstations, to expand the limited number of high-speed PCI Express (PCIe) lanes provided by the CPU. It acts as a traffic controller to enable robust peer-to-peer communication between multiple GPUs and other PCIe devices.
Without a PLX switch, connecting multiple GPUs to a motherboard can quickly exhaust the PCIe lanes coming directly from the CPU. This forces the GPUs to operate at a reduced number of lanes (e.g., x8 instead of x16), which limits bandwidth and performance. A PLX switch solves this problem by functioning as a PCIe expansion hub.
Enables multiple GPUs: A single PLX switch can connect a CPU's single high-bandwidth PCIe link (e.g., x16) to several GPUs simultaneously.
Improves GPU-to-GPU communication: In GPU-centric applications like deep learning, peer-to-peer (P2P) communication between GPUs is critical. A PLX switch can connect multiple GPUs to each other directly, which significantly reduces the latency that would occur if the data had to pass through the CPU first.
Maximizes bandwidth utilization: It dynamically manages the traffic and bandwidth among the connected devices. For example, if only one GPU is active, the PLX switch can grant it the full x16 bandwidth. If multiple GPUs are active, it intelligently distributes the available bandwidth to maximize throughput.
Deep Learning and AI training: AI workloads rely heavily on parallel processing across multiple GPUs. A PLX switch optimizes the communication required for models that need to share and transfer data rapidly between GPUs.
High-performance computing (HPC): In scientific and engineering simulations, compute clusters with several GPUs require high-speed, low-latency inter-GPU communication, which is facilitated by a PLX switch.
Multi-GPU workstations and servers: Systems with more GPUs than the CPU can directly support often rely on PLX switches. This includes 4-GPU, 8-GPU, and even higher-density servers.
Here are what the PEX89104 enables in your GPU server:
High Downstream Port Density: Many GPUs + NVMe devices can be connected without requiring multiple PCIe switches or motherboard PCIe slots.
Flexibility in Slot Allocation: The switch enables you to reassign or provision downstream ports (for example, some for GPUs, some for NVMe storage) without having to rewire upstream directly to CPU.
Cost / Space Efficiency: Instead of having multiple host links or separate I/O boards, one switch concentrates the interconnect, saves motherboard real estate, and helps with routing.
Scalability: More GPUs/devices can be added (depending on chassis / power / cooling) without redesigning the base motherboard.
Better Bandwidth Utilization: Because PCIe Gen5 has very high per-lane bandwidth, the switch helps in aggregating traffic efficiently. For instance, GPU-GPU, GPU-NVMe, or GPU-network paths can go through the switch, reducing reliance on CPU lanes (which are often precious).
The Hexadata HD-RS4000 Version : TUR-02/10G is an advanced AI and HPC platform that pushes the envelope of performance, efficiency, and scalability. By leveraging AMD's latest processors and NVIDIA's cutting-edge GPU technologies, the system is designed to accelerate workloads in AI, deep learning, data analytics, gaming and HPC environments. With a comprehensive array of features—including PCIe Gen5, and GPU memory bandwidth—the Hexadata platform is well-suited for organizations aiming to stay at the forefront of AI innovation while ensuring cost-effective and energy-efficient operations.
This solution is built for businesses and research institutions seeking scalable infrastructure to support the most demanding AI, data analytics, and high-performance computing tasks.
Reliability and Power Efficiency:
The HD-RS4000 Version : TUR-02/10G is designed with 4 x 3200 W 80 PLUS Titanium redundant power supplies, ensuring exceptional power supply efficiency and reliability. This feature guarantees that the system remains operational even during power interruptions, making it suitable for critical workloads.
80 PLUS Titanium certification ensures the system consumes less power, providing a more energy-efficient solution compared to previous-generation systems.
Scalable and Flexible Design:
The 4U rackmount form factor ensures that the system is well-suited for deployment in data centers, where space is often a concern but high computational power is essential. The form factor allows for easy scaling, while the system's robust design ensures reliability and longevity in demanding environments.
For management and maintenance of a server or a small cluster, users can use the Hexadata Management Console, which is preinstalled on each server. Once the servers are running, IT staff can perform real-time health monitoring and management on each server through the browser-based graphical user interface. In addition, the Hexadata Management Console also provides:
TPM 2.0 Module
For hardware-based authentication, the passwords, encryption keys, and digital certificates are stored in a TPM module to prevent unwanted users from gaining access to your data. Hexadata TPM modules come in either a Serial Peripheral Interface or Low Pin Count bus.
The Hexadata HD-RS4000 TUR-02/10G is a highly scalable, GPU-dense server platform engineered to deliver exceptional performance for the most demanding high-performance workloads. Built to excel in AI, HPC, deep learning, data analytics, and beyond, this server integrates the latest advancements in PCIe Gen5 switching technology with unparalleled GPU support.
Fully compatible with the latest NVIDIA Enterprise GPUs, the HD-RS4000 ensures seamless integration and optimized performance for NVIDIA's cutting-edge accelerators, enabling organizations to harness the full power of GPU computing.
Artificial Intelligence & Machine Learning : Maximize training and inference efficiency with NVIDIA's latest GPUs and high-bandwidth PCIe fabric.
High-Performance Computing (HPC): Deliver powerful parallel computing for simulations, modeling, and scientific research.
Data Analytics & Big Data : Accelerate data processing pipelines with high-throughput I/O and GPU acceleration.
Rendering & Visualization : Support complex rendering workflows and real-time visualization demands.
Hexadata's HD-RS4000 TUR-02/10G GPU Servers provide unmatched scalability and performance tailored for enterprise and research environments. Combining the advanced Broadcom PEX89104 PCIe switch and full compatibility with the latest NVIDIA Enterprise GPUs, this platform empowers your organization to push the boundaries of compute-intensive workloads with confidence and efficiency.
