Not logged in » Login
Jul 11 2017

Redesign Your Data Center with PRIMERGY and Intel® Xeon® Processor Scalable Family


Back in April, a so-called Product Change Notification from Intel suddenly appeared online, providing CIOs and IT departments with a first glimpse at the naming conventions and tech specs of the company's upcoming "Purley" line of server processors. Essentially a 'port' of the Skylake microarchitecture to the Xeon brand, it is intended to serve as the "foundation for secure, agile, multi-cloud data centers" and gives users a wider range of "workload-optimized" performance options to choose from. Starting in July, the Purley platform – officially dubbed the Intel® Xeon® Processor Scalable family – will appear in our new, fourth generation of dual- and quad-socket PRIMERGY servers as well as the PRIMEQUEST 3800B.

For quite some time, CIOs and their staff have been looking out for processors that would not only help them run ever more data-intensive workloads at maximum speed with low latencies, but also improve agility and security within one and the same single step. From their perspective, the ideal "CPU" family should be capable of handling all types of demanding tasks, from mission-critical OLTP workloads through real-time analytics to machine learning/deep learning routines that power applications as diverse as image recognition, natural language processing, CRM and recommendation systems or genome sequencing, all while being tamper-proof and cloud-ready. With its new Intel® Xeon® Processor Scalable family, Intel tries to cover as much of that ground as possible in what it has pitched as "biggest set of data center platform advancements in this decade."1 But what exactly does the new brand name stand for, and in which way will the upcoming chips differ from their predecessors?

Skylake + Lewisburg = Xeon Processor Scalable Family
Simply put, Intel's new chip family will replace the existing Xeon E5 and E7 processors that were previously used in multi-socket servers. While these could be tuned to accommodate specific customer needs, the idea behind the Intel® Xeon® Processor Scalable family is to offer more integrated functionality as well as a larger set of discrete expansion options for storage, networking and compute acceleration. To achieve this, Intel is bringing its 14 nm Skylake microarchitecture to the higher-level Xeons, joins it with C620 series ("Lewisburg") chipsets and connects both to the LGA 3647 socket that was first introduced last year alongside the Xeon Phi x200 accelerators. The new microarchitecture allows for moderately higher core counts (up to 28 max) and clock speeds (up to 3.6 GHz) compared to previous models; in addition, a fresh L1/L2 cache design enables lower latencies that result in faster, more responsive applications. By contrast, the chipset adds more visible improvements, for example new controllers that increase memory and PCIe I/O bandwidth by roughly 50% or support for DMI 3.0, which allows for connections to larger sets of up to 14 SATA 2 or up to 20 PCIe Gen 3 storage devices. The new socket for its part adds technologies such as Intel® UltraPath Interconnect (UPI) – think QPI on steroids – for faster data transfers between processors, and will support non-volatile memory. As a result, machines equipped with the new processors will be a better fit for environments with high VM densities, such as private or hybrid clouds, or workloads that demand extensive parallelization. According to Intel, this should allow for an up to 3.9x higher virtualized throughput and more VMs per server when compared to performance estimates for Intel® Xeon® E5 processors based on virtualization infrastructure consolidation workload.

Agility, Acceleration and Security
Most of the above are more or less expectable advancements that go along with every fresh processor generation that appears in the market, regardless of vendor. So in order to set the Intel® Xeon® Processor Scalable family apart from the rest, Intel decided to offer up a neat bundle of extras that make the processor platform more flexible, agile and secure:

  • Select Intel® Xeon® Processor Scalable family models will support Intel's AVX-512 instruction set, which debuted on the Xeon Phi x200 accelerators last year and increases the number of floating-point operations by 100% per core, which significantly speeds up large database and analytics applications. Likewise, the company's QuickAssist Technology (QAT) more than doubles encryption and compression speeds, which will be helpful in backup scenarios.
  • Moreover, AVX-512 will boost the performance of Intel's built-in protection mechanisms such as TXT (Trusted Execution Technology), PTT (Platform Trust Technology) or Boot Guard that deter attackers from directly tampering with the hardware. QAT and PTT also offer integrated protection of encryption keys. Finally, as a result of the switch to the Skylake microarchitecture, the new Xeon processors also inherit the MPX (Memory Protection Extension) functionality that aims to prevent buffer overflows and thus minimize a server's attack surface.
  • Intel's Resource Director Technology for SLAs in combination with its VMD for NVMe SSDs simplifies and improves cloud and storage management.
  • Finally, by adding its Omni-Path high-performance communications architecture to the mix, Intel manages to significantly boost PCIe connectivity, with bandwidths increasing up to 58 Gb/s in x8 HFI and up to 100Gb/s in x16 HFI setups.



Fig. 1: Intel's new color scheme will be present in product logos and make it easier for customers to pick adequate models from the large Xeon® Processor Scalable family


Availability and Dosage Forms
Intel's new Xeon® Processor Scalable family is available immediately as a core component of Fujitsu's 4th-generation PRIMERGY dual- and quad-socket servers as well as the eight-socket powerhouse PRIMEQUEST 3800B.

Just like previous Xeon generations, the new product line comes in a multitude of designs that differ in core counts and clock speeds; however, it should be pointed out that – thanks to the expanded set of integrated functionalities – customers will be able to choose from a much broader selection of almost tailor-made processors. Moreover, a larger number of processor models also means they get more freedom of choice regarding base units, RAID options, storage devices, etc. To help everyone find the processor that best fits their needs, Intel has introduced a new color scheme (cf. Fig. 1 above) that roughly indicates what users may expect from a given Xeon processor: Bronze models offer entry-level performance, Silver stands for "efficient performance at low power," Gold means they get great performance and fast memory plus more interconnects and accelerator engines, and Platinum is simply top notch in every aspect, including security and agility. It's plain to see that the color codes only allow for a first orientation, but help from Fujitsu is right around the corner: A CPU-Guide is downloadable here to make sure you match correctly when sizing your servers. Video footage from Intel's product launch is available here.

1. Cf. Lisa Spelman: „Re-Architecting the Data Center: The Intel Xeon Processor Scalable Family.” Available online at: Retrieved 2017-07-10.

Timo Lampe


About the Author:

Timo Lampe

Senior Specialist Marketing Manager – Data Center Systems / Server


Comments on this article

No comments yet.

Please Login to leave a comment.


Please login

Please log in with your Fujitsu Partner Account.


» Forgot password

Register now

If you do not have a Fujitsu Partner Account, please register for a new account.

» Register now