Meet the demands of Accelerated HPC, big data and I/O-heavy compute with an AMD server designed specifically for these tasks.

Until recently, data centers relied on cluster systems built from off-the-shelf servers with x86 processors and high speed networks. But CPU-based systems struggled to handle the processing needs of high performance computing (HPC), big data and deep learning (DL).

The problem is this: HPC, DL, IO-heavy HPC and big data computing workloads require distinct hardware designs to maximize the features that improve performance specific to the applications they are running. After all, you wouldn't use a coffee machine to make tea.

Matching Accelerated HPC workloads

Modern HPC-related workloads often consist of simulations that require the processing of large amounts of floating point calculations to simulate or model complicated processes. Such HPC simulations can be performed to model materials and molecular systems, weather forecasts and astronomy, fluid dynamics, financial markets, oil and gas, physics, bioscience and many other fields.



In these areas, HPC computing increasingly uses GPU-based accelerators for simulation and processing performance needs because they are designed to run massively parallel applications and process a single instruction across large amounts of data at the same time.

Meeting I/O-heavy system demands

IO-heavy computing is a different beast. It requires systems to efficiently read/write and store large amounts of data on disks. Crucially, if a process is limited by the speed of the IO subsystem, it can become IO Bound. This is common for database applications as they wait for data to be fetched from disk before the calculations are performed.

One solution is to use NVMe to provide much higher input/output operations per second (IOPS) performance than previous SATA storage devices. NVMe is an open logical-device interface specification for accessing non-volatile storage media attached via a PCI Express (PCIe) bus.

Applications with heavy read/write storage requirements can benefit from using NVMe storage because it is directly accessible to the processor.

Getting the most from big data

Big data involves analyzing, extracting information from and storing data sets that are too large or complex to be dealt with by traditional data-processing application software. So, for best performance it requires a server that is specifically designed for heavy-duty number-crunching.



Big data tasks include predictive analytics, user behavior analytics, or certain other advanced data analytics methods to extract value from data and then store it. This requires systems with a high processing speed and bulk storage, such as spinning disks, to process and store the huge volumes of information generated.

What to consider when selecting HPC infrastructure

Organizations should look at a number of factors when determining the infrastructure they require.

• Will the system be used for general data processing, HPC/DL, big data or IO-heavy workloads?
• Does the application require fast access to data storage? This requires an infrastructure containing many NVMe solid-state drives (SSDs).
• Does the application benefit from highly parallelized GPU acceleration? This requires an infrastructure with many GPUs.
• Does the application need a large memory footprint? This requires a system with maximum Dual In-line Memory Module (DIMM) slots.

What Tyan can do for you

The following Tyan servers with state-of-the-art 3rd Gen AMD EPYC™ processors are capable of meeting the most demanding processing needs of modern workloads. With a 7nm doubled core count from 32 to 64 cores per socket, these CPUs have the most cores, the most per-socket performance, the largest amount of memory, the most PCIe lanes, and the fastest PCIe lanes on the market today.

For Accelerated HPC Computation, Tyan's Transport HX TN83-B8251 is a 2U 4-GPU server that can handle HPC and DL applications and works well with other GPU accelerated workloads.

In IO-heavy HPC Computing scenarios, Tyan's Transport CX GC79A-B8252 server provides excellent support for organizations doing IO-heavy computing with a variety of memory-based computing applications. This is due to the large number of NVMe bays (up to 12) and 32 DDR4 DIMM slots.

And for big data computing, the Tyan Transport CX TN73-B8037-X4S is an ideal solution for big data workloads and high-density data center deployment that targets scale-out applications with large numbers of nodes. The TN73-B8037-X4S platform is specifically designed for HPC and CSP (Cloud Service Provider) front-end processing server applications.

Built for the most demanding technical computing workloads, the Tyan Transport series is a line of world class barebones server systems based on AMD’s latest technology, including 3rd Gen AMD EPYC™ processors with AMD 3D V-Cache™. For more information about all of our AMD products, view the Tyan/AMD Product Catalog 2022.