Intel® Select Solutions for HPC
The use of high performance computing (HPC) to find new insights and drive innovation is growing more pervasive in both academia and industry. Intel® Select Solutions for HPC identify systems that target the needs of HPC workloads and speed deployments for a fast return on investment in cluster resources—and accelerated time to insights.
Branded HPC solutions provide broad compatibility with HPC applications and compliance with the Intel® HPC Platform Specification. The specification combines standard industry practices, common HPC application requirements, and the Intel® software runtime environment to provide a common foundation for Intel® architecture-based solutions.
Intel® Select Solutions for HPC take the guesswork out of selecting the right set of high performance computing products, hardware configurations, and software. Platforms must include a minimum set of components that provide optimizations for HPC and must exceed minimum performance thresholds that are key to scaling application performance.
These solutions must also have a proven deployment path to speed acquisition and time to first use. Branded solutions mean the combination of compute, fabric, memory and storage, and software meets HPC application needs with validated steps to build and deploy the solution.
The Intel® Select Solutions for HPC family provides solutions targeting specific needs.
- A base solution for simulation and modeling targeting broad commercial and community HPC codes.
- Simulation and visualization adding cutting-edge capabilities to spread visualization workloads across the HPC cluster.
- Genomics combining data analytics and genome sequencing software from the Broad Institute on top of a base HPC solution.
- HPC/AI convergence solutions enabling traditional HPC and emerging artificial intelligence workloads on a unified platform.
Intel® Ingredients for HPC
Intel® technologies span the spectrum of HPC, from workgroup to supercomputing. It takes a sophisticated set of ingredients to meet application needs, including compute, fabric, memory, and storage components. Intel offers a comprehensive portfolio of HPC hardware to meet the range of needs in HPC.
Compute: Intel® Xeon® Scalable processors
Intel® Xeon® Scalable processors provide a foundation for breakthrough HPC systems, plus a leap in agility and scalability. By helping maximize vector floating point performance and efficiency in HPC systems, these processors deliver high performance with fewer servers.
The Intel® Xeon® Scalable processor family was designed to balance the compute, memory, storage, and networking needed for HPC systems—all while supporting massive scalability to tens of thousands of cores. It features an improved core count and mesh architecture, as well as support for the latest memory and fabric technologies. All this helps maximize performance across resources without creating bottlenecks.
Intel® Xeon® Scalable processors have built-in optimizations for both HPC and AI. These include features like Intel® AVX-512 for accelerating vector processing and Intel® Deep Learning Boost (Intel® DL Boost) to enhance inference performance. With a common platform for HPC, AI, and data analytics, enterprises are better prepared to converge these practice areas and maximize the utilization of their HPC hardware.
Acceleration: Intel® FPGAs
A field programmable gate array (FPGA) is a semiconductor-integrated circuit that can be reprogrammed even after the hardware has been deployed. Intel® FPGAs can be customized to accelerate HPC workloads, helping engineers adapt to evolving standards and requirements.
Fabric: Intel® Omni-Path Architecture (Intel® OPA)
Intel® Omni-Path Architecture (Intel® OPA) is designed to scale out for extremely large clusters, cost-effectively. The Intel® OPA product line includes PCIe* adapters, silicon, switches, cables, and management software built upon proven Intel® technology. This provides low-latency interconnect without sacrificing performance or reliability.
Fabric performance isn’t just for large clusters, though. Intel® Omni-Path Architecture (Intel® OPA) provides key technologies to power small and medium clusters for application performance and is a key element helping keep data moving in high-performance file systems.
Memory and Storage: Intel® Optane™ DC Technology
To enable the next generation of HPC, system architects will need to rethink memory and storage. Intel® Optane™ DC technology is a revolutionary approach that bridges gaps in the storage and memory hierarchy.
Supported by 2nd Generation Intel® Xeon® Scalable processors, Intel® Optane™ DC persistent memory provides up to triple the maximum storage per node for larger in-memory data sets, while reducing data latency below hard disk drives (HDDs) or solid state drives (SSDs). Persistence helps ensure that large models running in off-peak hours will still be there even with a power surge or system maintenance.
Intel® Solid State Drives are helping system architects create a new layer between the data pipeline and storage resources. Intel® Optane™ DC SSDs are extremely responsive under any load and highly predictable. Intel® 3D NAND SSDs provide greater performance, higher density, and higher reliability than hard disk drives, and are also on a path to approach cost parity with hard disk storage.
Network: Intel® Ethernet Products
Networking components play a critical role in helping eliminate bottlenecks in HPC systems. Intel® Ethernet network adapters, controllers, and accessories bring enhanced agility to HPC systems with cost efficiency and the broad compatibility of Ethernet. These products provide a good range of fabric performance, allowing solutions flexibility while providing standard connections for devices and technologies to drop into designs.
Preparing for the Next HPC Breakthroughs
With outstanding processing performance, memory bandwidth, and scalability, Intel delivers a powerful platform for all types of HPC systems—from small clusters to supercomputers. With our latest innovations across compute, fabric, memory, and storage, Intel is helping organizations be ready to achieve world-class HPC system performance at unprecedented scale.
Intel® Xeon® Scalable processors were designed to balance compute, memory, storage, and networking in HPC systems—all while supporting scalability to tens of thousands of cores.