High Performance Computing and Visualization
Ground-breaking Unified Platform
Simulation and Visualization
Powered by Intel Rendering Framework
Faster and Deeper Understanding of Large Data Sets
Faster and Deeper Understanding of Large Data Sets
Intel® Select Solutions for HPC offers an easy and quick-to-deploy infrastructure that reduces the complexity of advanced computing and helps accelerate the time to actionable insights for users in industry and science. The portfolio of solutions includes workload-optimized hardware configurations with pre-loaded software for simulation and visualization. Intel® Select Solutions for Simulation & Visualization quickly deploy an optimized infrastructure that delivers faster simulations and photorealistic visualizations that are crucial in the age of big data thorough Atipa and Intel® Select Solutions.
Visualization plays a key role in understanding complex patterns and their dependency on underlying models in science and engineering. The current High-Performance Computing (HPC) paradigm uses large-scale compute clusters for simulation and modeling but does not address the need for post-simulation visualization. Instead, data produced by simulations is copied to high-end workstations and servers with GPUs for visualization and rendering. As simulations continue to grow, moving and copying data becomes increasingly inefficient and time consuming.
Software Defined Visualization (SDVis) is an open-source initiative led by Intel to use CPUs rather than GPUs for high-fidelity rendering and visualization. The initiative has materialized in the development of three high-performance, open-source software rendering libraries optimized for 2nd Gen Intel® Xeon® Scalable Processors: the Embree Ray-Tracing Kernel Library, the OSPRay Distributed Ray-Tracing Framework, and the OpenSWR OpenGL-Compatible Software Rasterizer. Existing applications and visualization tools such as ParaView, VisIt, Corona, and 3DExcite, to name but a few, can utilize these software libraries to improve the visual fidelity and performance of scientific visualizations and professional rendering.
The motivation behind SDVis is simple: high-accuracy simulations produce data sets too large to fit in the on-board memory of GPUs. Rendering large data sets on GPUs therefore requires multiple tiling passes and data transfers to and from the GPUs via the PCIe bus. In contrast, CPU-based rendering has access to hundreds of Gigabytes of memory in a single server to hundreds of Terabytes of memory in clusters of servers. As a result, SDVis can produce more accurate and more realistic images in much shorter time.
Eradicating the need for targeted graphics hardware allows scientists and engineers to use a single unified platform for simulation, modelling, and visualization. With the price tag of a single GPU approaching the price of an entire HPC node, the ensuing cost savings free up budget for additional multi-purpose cluster nodes suitable for both computation and visualization. Converging HPC and SDVis in a single unified platform furthermore improves resource utilization, power efficiency, and scalability.
Key Advantages of Software-Defined Visualization
Visualizations can be rendered on the same HPC clusters where data is generated without first having to copy massive data sets to GPU workstations or rendering farms.
Large Memory Capacity
CPU-based rendering has access to the aggregated memory pool of all servers in HPC clusters. This allows visualization of much larger data sets with much improved accuracy and fidelity.
The ability to use the memory capacity and parallelism of entire clusters of 2nd Gen Intel® Xeon® Scalable Processors offers huge performance gains over specialized graphics hardware limited by memory capacity and PCIe bandwidth.
Larger HPC clusters allow more accurate and more detailed simulations producing larger data sets. SDVis enables visualization-driven discovery to grow with the simulations.
CPU-based visualization eliminates the need for expensive graphics hardware. A single unified system for simulation, modeling, and visualization increases ROI by decreasing acquisition costs and speeding-up time to discovery.
Intel® Select Solutions for Simulation and Visualization
Intel® Select Solutions for Simulation and Visualization provide a fast path for purchasing and deploying a cluster to enable visualization workloads without the cost, limitations, and system challenges of GPU-based solutions. Intel® Select Solutions for Simulation and Visualization consist of a pre-validated selection of software and hardware components designed to meet the demands of HPC and enterprise applications that include visualization workflows. These systems also provide the capabilities and agility needed to eliminate the need for multiple single-purpose systems.
Intel® Select Solutions for Simulation and Visualization are verified solutions that combine Intel technologies with the Intel® Scalable System Framework (Intel® SSF) to help eliminate the need to perform lengthy research to determine the right hardware and software for maximizing both simulation and visualization capabilities. The solutions deliver optimized performance and compatibility for simulation and visualization applications registered with Intel® SSF.
Intel® Select Solutions for Simulation and Visualization define performance watermarks that demonstrate optimized capabilities for HPC and visualization applications. Verified solutions meet or exceed Intel design and testing standards, including two visualization-focused benchmarks (OSPRay and ParaView), two well-known compute application benchmarks (HPL and HPCG), and three well-known computing micro-benchmarks (DGEMM, STREAM, and IMB PingPong) that support a range of important system aspects and indicate potential scale-up and scale-out performance.
Fast-Track Simulation and Visualization with Atipa Polaris Select Clusters
Atipa Technologies’ Polaris Select clusters offer a scalable, unified platform for High-Performance Computing and Visualization (HPCV). They are pre-configured, turn-key solutions including all necessary software for parallel computing and cluster management (compilers, MPI, MKL). The hardware and software specifications comply with the reference architectures for both Intel® Select Solutions for Simulation and Visualization and Intel® Select Solutions for Simulation and Modeling.
Atipa Polaris Select HPCV clusters spotlight the latest Intel technologies for high-performance computing, high-bandwidth low-latency fabrics, and high-throughput storage:
2nd Gen Intel® Xeon® Scalable Processors feature significant enhancements that benefit HPC and SDVis applications, including improvements in I/O, memory, and Intel® Advanced Vector Extensions 512 (Intel® AVX-512).
Intel® Omni-Path Architecture (Intel® OPA) provides 100 gigabits per second (Gbps) unidirectional bandwidth and a low-latency, next-generation fabric.
Intel® Non-Volatile Memory Express (NVMe) SSDs outperform SAS/SATA SSDs by an order of magnitude, increasing system responsiveness and accelerating I/O for data-intensive workloads.
Atipa Polaris Select HPCV clusters are fully customizable solutions that are pre-tested and validated with Intel® Cluster Checker to follow the Intel® Select Solutions reference architecture. Proof-of-concept configurations start from a single head node with three unified HPCV nodes providing an aggregated memory capacity of 768GB, the equivalent of twenty-four high-end GPUs. The recommended minimum configuration for production systems consists of a single head node and eight unified HPCV nodes. Table 1 shows the minimum hardware and software specifications satisfying the Intel® Select Solutions for Simulation and Visualization and Intel® Select Solutions for Simulation and Modeling reference architectures.
Atipa Polaris Select HPCV clusters draw upon the power of 2nd Gen Intel® Xeon® Scalable processors, Intel® Omni-Path Fabric (Intel® OPA Fabric), and Intel’s software defined visualization libraries to help researchers, engineers, digital content creators, and others gain a deeper understanding of data or deliver stunning photorealistic imagery. Atipa Polaris Select HPCV clusters make high-fidelity, photorealistic images possible, even as data sets become increasingly large and more complex.
INTEL® SELECT SOLUTIONS FOR SIMULATION AND VISUALIZATION BASE CONFIGURATION
DATA DRIVE RAID CONTROLLER
*Minimum hardware and software specifications for Intel® Select Solutions for Simulation and Visualization.
MINIMUM PERFORMANCE STANDARDS
*Verified to meet or exceed the above minimum performance capabilities
Atipa HPC Products and Solutions: https://www.atipa.com
2nd Gen Intel® Xeon® Scalable Processors: https://www.atipa.com/2nd-gen-intel-xeon-scalable-procs
Intel® Select Solutions: https://www.intel.com/selectsolutions
Software Defined Visualization: https://software.intel.com/sdvis
Intel® Omni-Path Architecture: https://www.atipa.com/intel-omni-path
Intel® Parallel Studio XE: https://www.atipa.com/intel-parallel-studio-xe
Intel® Scalable System Framework: https://www.intel.com/ssf