NEC HPC LX Series

AMD

SCALE YOUR DATA CENTER WITH EFFICIENCY

Whether you’re running a private cloud or a cluster within your data center, virtualization is a critical part of any modern IT infrastructure.By consolidating hardware resources and streamlining IT management, hypervisors such as VMware ESXi™ allow enterprises to run efficiently and scale as needs arise. 4th Generation AMD EPYC™ processors support virtualized environments with the performance, energy efficiency, and value required for virtualized IT infrastructure workloads.AMD EPYC 9004 Series processors are offered in a variety of core-count options—from 16 to 128 cores—that can deliver more vCPU density, energy efficiency, and performance per dollar value than 4th Gen Intel® Xeon® Scalable CPU offerings. By using fewer servers and less power to deliver similar performance in your data center, your enterprise can shrink its related carbon footprint without sacrificing performance. For example, 11 two-processor (2P) AMD EPYC 9684X processor–based servers (with 96 cores/CPU) can deliver similar VMware® VMmark® performance to 20 2P Intel® Xeon® Platinum 8940H processor–based servers (with 60 cores/CPU), along with 39% estimated lower energy costs and estimated annual savings of 48.7 metric tons of CO2 equivalent (MTCO2e).AMD compiled and compared matched-pair scores for a variety of 3rd and 4th Gen AMD EPYC processors and Intel’s Top of Stack 4th Gen Xeon Scalable processor running the VMware VMmark 3.1 benchmarks. See Figure 1 to review the results of these analyses, and then make your decision about what processor to use based on clear data. More on AMD EPYC 

AMD Instinct™ Accelerators: Technology

Immense computational power coupled with the fusion of high-performance computing (HPC) and artificial intelligence (AI) is enabling researchers to tackle grand challenges once thought beyond reach, from climate change to vaccine research.Accelerating these workloads is possible by adding graphics processing units (GPUs) to systems. The GPUs work in tandem with CPUs to speed computation for these graphics-intensive general-purpose applications and scientific workloads. AMD Instinct™ MI350 Series deliver leadership performance for the data center at any scale, from single-server solutions to Exascale-class supercomputers. They are uniquely well-suited to power demanding AI and HPC workloads, offering exceptional compute performance, large memory density, high bandwidth memory, and support for specialized data formats. More on MI350

NEC HPC LX Series

NEC Standard HPC Platform

NEC offers a wide variety of server platforms. For HPC, dense dual socket x86 compute nodes with an HPC interconnect have proven a successful concept for the main HPC applications.
With four servers in a 2U chassis costs and space per server is minimized and the flexibility of configuration is increased compared to larger blade solutions.

NEC Large Memory HPC Platform

With the increase of core count and memory capabilities even in dual socket nodes even a 1U or 2U system can provide applications with the needed scalability and space. For highest demands NEC can also provide servers with up to 8 CPU sockets, providing an even larger memory configuration and SMP solution.

NEC Advanced HPC Platform

The NEC SX-Aurora TSUBASA platform integrates seamlessly into the NEC LX Series. For advanced applications where vectorization is key, a selection of NEC SX-Aurora TSUBASA products are available.

To meet the needs of specialized applications NEC offers different types of accelerators. Here advanced users can find solutions that support specially developed applications. NEC supports the Intel and NVIDIA range of GPU accelerators and have experience in setting up demanding setups with many accelerators.

NEC Infrastructure Nodes

Operating an HPC cluster does not only mean running computational jobs. For operational and maintenance tasks, monitoring and user control, NEC provides price efficient solutions with high flexibility. In addition, a cluster in production must be available and useable at all times. Therefore, we have developed different high availability concepts to make sure that a cluster is always available to its users. The solutions range from a simple active-passive server configuration to solutions with shared hardware storage to provide best in class failover capabilities. 

NEC HPC LX Series

HPC Infrastructure

From working with different datacenters, NEC have extensive experience to provide the best infrastructure solutions for the HPC environment. With the power consumption going up in newer generations of systems and the need for environmentally friendly cooling rising, customers notice that infrastructure is playing a larger role.

Indirect Cooling with Water Cooled Racks

With regards to the tradeoff between investing in infrastructure and investing in servers, the side-cooler solution is a highly interesting concept. Here, the airflow circulates from the hot server rear side through high performance side cooling elements back to the server front. By restricting the air volume to the rack a more effective cooling can be realized than by conventional air conditioning. Inlet water temperature of up to 28 degrees Celsius can be used.

This method is compatible with all standard servers and can even be combined with free-air cooling methods, which exploit low outside temperatures to cool the servers in an environmentally friendly way. With the help of additional cooling equipment one can also cover the peaks in summer. Many datacenters successfully use this model.

Direct Liquid Cooling

More and more datacenters put a lot of efforts into reducing their cooling costs even further by making the leap to direct hot-water cooling. In this concept the water has an inlet temperature of up to 45 degrees Celsius and directly cools CPU, memory and other equipment by elements that are part of a closed water circuit. NEC works together with several vendors in the field to enable the best solution for each datacenter.

• DLC Technology Whitepaper