The strength: from small to large

Prepare your business for the future by taking advantage of innovative and tested cloud hardware. RECT™ servers on Intel’s scalable Purley platform equipped with powerful Xeon® scalable processors bring the latest server technologies from state-of-the-art computing centres directly to your company. You attain a considerable leap in performance and agility in the various application scenarios. You bank on an entirely new performance class with sizable performance increases in the processor, RAM and storage, as well as in all I/O usages. At the same time, you bet on the security benefits of the most innovative of server technologies – likewise derived from the requirements of state-of-the-art computing centres as well as the high-level scenarios of deep learning and artificial intelligence.

Customised modular concept /
What Intel® calls “scalable” turns into a modular concept in the RECT™ server. As always: everything is tailor-made to meet the individual needs. But: everything from single and dual sockets to quad or octo sockets comes on a single platform. And: it features the perfect quality for every server scenario, from small office servers to computing clusters for scientific applications. The advantages are not only relevant for the level of performance, the new platform also realises a significant improvement in the price-performance ratio. All in all, the new RECT™ servers deliver a tailor-made and needs-based solution for all tasks to attain greater speed, more stability, higher efficiency and a substantially better price level.

Tremendous performance characteristics
RECT™ servers on an Intel® Purley platform with Intel® Xeon® scalable CPUs provide outstanding performance packages for your application scenarios. In coordination with the CPU performance classes (from bronze to silver and gold to platinum), RECT™ servers likewise always supply precisely the level that delivers a coordinated yet powerful, an efficient but price-sensitive overall package to the company.

Always in a league of its own
RECT™ servers with Intel® Xeon® scalable processors with single and dual socket servers consistently deliver a bandwidth of 4 to 56 cores, that is, 4 to 112 threads per server, offering the connection of 48 PCI-E 3.0 slots for controllers, NVMe storage pools and network cards for each CPU. The new cache architecture within the CPU now features a larger L2 cache and a “non-exclusive” L3 cache, to attain a considerably better cache performance. As well, the optimised Turbo Boost technology of the CPUs enables better performance with higher clock speeds up to 3.7 GHz. The RAM now runs at clock speeds of up to 2600 MHz and is connected by way of 6 channels, thereby achieving a performance boost by up to 60 percent. The AVX-512 extension designed to accelerate floating point operations also delivers a greater computer throughput relative to the energy consumption or the clock rate. It doubles the computing speed, particularly for computation-bound applications.

Key facts about the platform

• standardised CPU socket LGA3647 for single, dual, quad and octo socket servers
• CPUs divided into 4 performance levels: Bronze, silver, gold, platinum
• CPUs with 4 to 28 cores and 4 to 56 threads with optimised CPU cache performance (larger L2 and ”non-exclusive” L3 cache)
• new CPU Interconnect Intel® UPI instead of the previous Intel® QPI for a faster communication between the CPUs with bit rates of 9.6 to 10.4 GT/s
• Intel® AVX-512 and Intel® Omni-Path for performance optimisation in the computing and transmission speed
• Intel® VMD and Intel® VROC for hot-plug U.2 NVMe SSDs and general NVMe-RAID control
• 768 GB of DDR4-RAM per CPU, connected by way of 6 memory channels
• 48 PCI-E 3.0 slots per CPU

RECT™ servers with Intel® Xeon® processors ultimately deliver a targeted optimisation for nodes. When compared with the previous generations of a normal product lifecycle, they are now capable of performing up to 4.2 more MVs and up to 5 times more transactions per second. What’s more, IOPS are 5 times faster, while the latency is 70 percent lower.