Opening the future of high performance computing and artificial intelligence with revolutionary accelerated computing, presenting the HPE Apollo 6500 Gen10 Plus System. This groundbreaking solution delivers highly scalable performance and will revolutionize the use of high performance computing and artificial intelligence.
In today’s digital economy, data drives innovation. Intelligence changes the way we think, interact and understand the world. The ability to use reliable predictive data is the difference between success and failure for businesses today.
The demand for high performance computing and artificial intelligence is skyrocketing as organizations rush to get their data to work. To achieve this goal, a number of industries in Russia are seeing a significant increase in the number of deployments of solutions for high performance computing and artificial intelligence in order to increase productivity and gain a strategic advantage in a highly competitive market. Segments such as financial services, healthcare and life sciences, manufacturing, automotive and retail around the world are undergoing a complete transformation as organizations invest in advanced capabilities and tools to help them work faster and smarter.
Hybrid (using dedicated GPUs) computing delivers highly scalable performance and faster, more reliable computing power to optimize the most demanding HPC and AI workloads. HPE recently announced a groundbreaking addition to the HPE Apollo portfolio that takes hybrid computing to the next level. System HPE Apollo 6500 Gen10 Plus Is the next generation in this product line that will take the use of high performance computing and artificial intelligence to a new practical level.
The HPE Apollo 6500 Gen10 Plus System is specifically designed to deliver great value to customers:
the system allows for faster analysis to gain a competitive advantage, allowing better use of data;
the system has the best performance due to the close connection of processing power with the best GPUs in the industry to provide maximum performance for the tasks being solved;
Fully tested and tuned by HPE certified by NVIDIA for demanding workloads of high performance computing, artificial intelligence, machine learning and deep learning;
The HPE Cray Supercomputer architecture now includes the HPE Apollo 6500 Gen10 Plus system as a node in its air-cooled version, combined with HPE Slingshot interconnect and supported by the HPE Cray Operating System.
The HPE Apollo 6500 Gen10 Plus system harnesses the processing power of the latest NVIDIA A100 Tensor Core GPUs and 2nd Gen AMD EPYC processors to reduce the burden on legacy infrastructure and expand opportunities for data-driven innovation. NVIDIA A100 GPUs deliver low latency, high bandwidth to enhance these powerful hybrid computing solutions. With third-generation tensor cores, the NVIDIA A100 can scale efficiently to thousands of GPUs or, with NVIDIA Multi-Instance GPU (MIG) technology, can be split into seven isolated GPU instances to accelerate a variety of workloads. NVIDIA A100 offers significant improvements in tensor cores and double precision numbers, breaking numerous AI performance records and delivering up to 4.2x faster processing speed than previous generations of accelerators.
The HPE Apollo 6500 Gen10 Plus System maximizes GPU utilization (up to 16 GPUs per server) for fast data collection, mining, and insight, regardless of the demands of HPC and AI workloads. NVIDIA NVLink seamlessly connects GPUs so they can work together as one powerful accelerator. The NVLink Inteconnect provides a dedicated communications link that allows memory data to be transferred between GPUs. One A100 accelerator supports up to 12 NVLink connections for a total bandwidth of 600 GB / s.
2nd Gen AMD EPYC processors offer massive bandwidth and high core counts to continuously handle data-hungry GPUs. Integrated with NIVIDIA Mellanox HDR InfiniBand, these high-frequency processors add up to 200 Gb / s of bandwidth for every two GPUs, so even productive business systems running in a cluster can exchange data at twice the speed.
For the Apollo 6500 Gen10 Plus system, HPE introduced:
the first server optimized for 4 GPUs that delivers the best price ever for HPC;
an updated dual-socket AMD-based server that can support up to 16 PCIe GPUs, double what HPE has supported in the past.
In addition, the HPE Apollo 6500 Gen10 Plus System offers extensive storage options including up to 16 front-access storage devices, SAS / SATA SSDs, and up to six NVMe drives for faster access to data with less power consumption. HPE plans to complement these capabilities with new solutions that will include 16 NVMe drives for up to 6x the throughput. HPE customers can now leverage the power, frequency, and processing power of hybrid computing to dramatically reduce the time it takes to get insights.
The HPE Apollo 6500 Gen10 Plus System is designed with density and flexibility in mind to support a wide range of HPC and AI applications. Our solutions have long been recognized in the industry for their high memory bandwidth, flexibility and security to ensure fast data transfers.
Created for the era of ex-scale computing, HPE Apollo 6500 Gen10 Plus Systems further increase performance for the most demanding HPC and AI workloads with NVIDIA HGX A100 GPUs with Tensor Cores and NVLink interconnect or AMD Instinct MI100 with Gen 2 Infinity Fabric Link. The total performance can exceed 100 Teraflops (FP64). The ability to install one or two processors improves the balance of processor cores, memory, and I / O. Increased system flexibility is also achieved with support for 4, 8, 10 or 16 GPUs and a wide range of operating systems and options offered, moving closer to customized design, which means lower costs for customers, increased reliability and better serviceability.
In addition to the NVIDIA HGX A100 and AMD Radeon Instinct MI100 accelerators, there is a wide selection of PCIe GPUs for high performance computing or artificial intelligence.
The HPE Apollo 6500 Gen 10 Plus System can contain the following server types:
Up to two HPE ProLiant XL645d Gen10 Plus Servers – This is a 1-socket server that supports 4 NVIDIA HGX A100 or 4 AMD Instinct MI100 accelerators or 4 x double-wide PCIe or 8 x standard-width PCIe accelerators;
One HPE ProLiant XL675d Gen10 Plus Server is a 2-socket server that supports 8 NVIDIA HGX A100 or 8 AMD Instinct MI100 accelerators, or 10 double-wide PCIe accelerators or 16 standard-width PCIe accelerators.
Various high-speed interconnect fabrics are also supported – Ethernet, HDR InfiniBand and HPE Cray Slingshot. Direct liquid cooling (DLC) is available as an option to improve efficiency and power density, and is shipped from HPE factories pre-filled, fully integrated, rack-mounted and ready to plug into the water mains, helping to reduce commissioning times and reduce the cost of ownership , an increase in cooling efficiency and a higher density in terms of electrical power.
High-performance components require the same high-performance power and cooling solutions to provide full redundancy for top-end processors and accelerators up to 500W to keep your system ready for current and future business challenges.
Service-friendly and upgradeable solution with full power redundancy, readily available modular design, hot-swappable dual rotor fans, and interconnect factory with rear cabling are fully compatible with standard 1075mm deep server rack for fast and efficient deployment … A wide variety of operating systems are also available, including the HPE Cray Operating System, Microsoft Windows Server, Ubuntu, Red Hat, or VMware.
The HPE Apollo 6500 Gen10 Plus System uses HPE iLO5 Hardware Root of Trust and AMD Secure Processor, a dedicated processor for enhanced security embedded in the AMD EPYC processor (system-on-a-chip, SOC). Structurally, the server occupies 6U and is made on a modular basis. This design will allow flexible configuration of the system now and provide an opportunity for future upgrades.
The server board allows you to install AMD EPYC processors with thermal pack up to 280W and has 32 slots for DDR4-3200 RAM. Internal storage supports up to 16 SFF drives or up to 6 SFF NVMe drives, easily accessible from the front of the server. PCIe 4.0 enables a variety of high-speed computing fabrics, including Infiniband HDR (200 Gbps) and HPE Cray Slingshot.
The Apollo 6500 Gen10 plus system contains technologies from an ex-scale era – an era in which system performance is measured by Exaflops (quintillion floating point operations). HPE is already building several Exascale systems in the US. To build these systems, it was necessary to revise all the approaches that were applicable to supercomputers in order to overcome the “super-” and go beyond – to the scale of “Aix”. We have developed products for these supercomputers and now some of the technologies that we have included in these products have become available for the corporate market, including in Russia. You can learn more about these products and technologies, as well as their applicability in corporate data centers, at the Fostering Innovation and Insights on Demand webinar, which will take place on March 24 at 10:00 Moscow time. You can register for it by this link…