After recent introduction of ultrahigh-dense energy efficient RSC Tornado solution based on 3rd Gen Intel Xeon Scalable processors and Intel Optane 200 series persistent memory, a number of RSC’s customers got the opportunity to test the new platform with their HPC workloads. They include Joint Institute for Nuclear Research (JINR), Siberian Supercomputer Center (SSCC) of Siberian Branch of the Russian Academy of Sciences (SB RAS) based at the Institute of Computational Mathematics and Mathematical Geophysics (ICM&MG), Physical and Technical Institute named after A.F. Ioffe of RAS (Ioffe Institute).
Double distributed computing and big data processing speed up in high energy physics
JINR researchers used GRID technologies for high energy physics computations. They evaluated processing speed of CERN's big data and the speed of generation and reconstruction of model data in planned experiments to research quark-gluon plasma using JINR’s new particle accelerator named NICA. Testing showed that data processing speed achieved with new RSC Tornado solution based on 3rd Gen Intel Xeon Scalable processors and Intel Optane 200 series persistent memory is 1.93 times higher than with 2nd Gen Intel Xeon Scalable processors.
Computing speed of heat process modeling in impulse cryogenic lattice increased by 2 times as compared to previous processor generation. Quantum computing with QuEST simulator was accelerated by 33% on 32 qubits.
16x acceleration of star formation process simulation
Scientists of ICM&MG SB RAS achieved 1.52x acceleration of star formation process simulation with parallel implementation and use of new code for Poisson's equation with RSC Tornado solution based on Intel Xeon Platinum 8368Q processors as compared to previous generation Intel Xeon Platinum 8268 processors. Performance of auto-vectoring with AVX512 instructions and Intel oneAPI on Intel Xeon Platinum 8368Q processors increased by 16.1 times as compared to similar processor cores without auto-vectoring.
Modeling astrophysical objects with extreme energy output
Calculation of rotational distortion interaction with accelerated solar wind particles done by researchers of Ioffe Institute, demonstrated 76% acceleration as compared to previous processor generation.
RSC Tornado based on 3rd Gen Intel Xeon Scalable and Intel Optane 200 series
New generation of RSC Tornado solution is oriented on a wide range of demanding scientific research workloads and applied tasks. The updated portfolio of integrated software-defined and reconfigurable solutions targets classic HPC systems, efficient storage and data processing and will help create AI/ML/DL (Artificial Intelligence, Machine Learning, Deep Learning) systems. RSC Tornado solution provides the industry highest x86 architecture computing density of 967.45 Tflops per rack (+37% vs. previous generation), 2.45PB distributed RSC Storage on-Demand per rack (+36% vs. previous generation) with 3.67 TB/s IO bandwidth (2x vs. previous generation) and leading energy efficiency with 100% “hot water” liquid cooling of all electronic components, horizontal scalability from small systems with a few servers to huge clusters or server farms containing many thousands of server nodes. RSC Tornado solutions provide additional cost optimization benefits through support of open standards and new Intel server products.
RSC Tornado solution based on 3rd Gen Intel Xeon Scalable processors (up to 40 cores, TDP 270W), Intel Optane persistent memory 200 series, Intel SSDs and high-speed fabric 200 Gb/s has leading footprint and computing density (up to 153 nodes in one standard 42U rack), high energy efficiency and provides stable operation of computing nodes in “hot water” mode with cooling agent temperature up to +65°C at inlets of switching nodes and interconnects. Operation in “hot water” mode enables all-year free cooling (24x365) using only dry coolers running at ambient air temperature up to +50°C, and complete elimination of chillers. As the result, average power usage efficiency factor (PUE) is less than 1.04, which is an outstanding score for HPC industry.
RSC Storage on-Demand solutions support NFS/Lustre/DAOS storage systems in one rack. New Intel open source DAOS (Distributed Asynchronous Object Storage) storage system provides the highest speed of data access with various patterns. This solution is optimal for AI (ML/DL) segment. It enables multi-layered storage based on Lustre file system in Disaggregated Composable Infrastructure and flexible management of NVMe disk pools, and enhances these levels with high-performance components based on DAOS. RSC uses its experience of building composable disaggregated solutions in DAOS management with RSC BasIS orchestration platform with the new and much more effective user interface (UI).
3rd Gen Intel Xeon Scalable processors
Optimized for HPC workloads as well as for cloud, AI, enterprise, network, security and IoT applications 3rd Gen Intel Xeon Scalable processors, based on Intel’s 10nm process technology, feature up to 40 powerful cores and a wide range of frequencies, feature and power levels. New benefits include PCI Express 4.0 (PCIe-Gen4) support, increased memory bandwidth, memory capacity up to 6TB per processor/socket with Intel Optane persistent memory 200 series and additional Intel AVX-512 instructions.
New Intel server processor family is the only data center CPU with built-in AI acceleration, end-to-end data science tools, and a vast ecosystem of smart solutions. The platform includes Intel SGX (Intel Software Guard Extensions) which protects data and application code in real time from the edge to the data center and multi-tenant public cloud, enabling enhanced collaboration using shared data – without compromising privacy. 3rd Generation Intel Xeon Scalable processors deliver a 50-60% competitive lead for key HPC life sciences, financial services and manufacturing workloads.
Intel Optane persistent memory 200 series
The 3rd Gen Intel Xeon Scalable platform supports the latest Intel Optane persistent memory 200 series delivering up to 6TB total memory per socket, 32% higher memory bandwidth vs previous generation. Intel Optane persistent memory is a new category of devices that sits between memory and storage. It has the flexibility to augment memory capacity beyond what is available on DRAM-only systems, or as a new persistent memory tier delivering high performance and low latency for accelerated data processing. Intel Optane persistent memory 200 series is the next-generation persistent memory module and helps users to extract more value from larger datasets, with a unique memory tier that combines large capacity and native persistence. Intel Optane persistent memory 200 series supports Enhanced Asynchronous DRAM Refresh (eADR), a new platform feature that can save volatile data residing in DRAM and CPU caches in the event of a power failure, preventing data loss without expensive battery-based memories.