Intel® Xeon® Scalable Processor
Intel has released their new family of Skylake Xeon processors code named Purley, officially referred to as Intel Xeon Processor Scalable Family which will be defined by a color code system that begins with the entry level Bronze (Xeon Bronze 31XX) series with up to 14 cores/28 threads and extends up to the high-end Xeon Platinum (Xeon Platinum 81XX) series that offers up to 28 cores/56 threads.
The new processor family brings about many changes in technology that make this the most significant change to Xeon’s since the release of Haswell over five years ago. The Xeon Processor Scalable Family includes features such as a mesh design to interconnect the multiple cores and L3 cache while reducing latency. The mesh design which is more efficient for connecting a higher number of cores is similar to that which is used for the Xeon Phi and replaces the ring interconnect that was used on previous generations of Xeon processors.
Another great improvement is the addition of VMB (Intel’s Volume Management Device) which resolves many of the previous issues of using NVMe in RAID configurations. With VMB the functionality of NVMe SSD is enhanced by allowing support for hot swap, LED error notification, and the ability to use multiple drives in a RAID configuration.
Other advances in the platform include Intel Quick Assist Technology which is a purpose built IC built into the Purley motherboard chipset that handles encryption, compression, and decompression of files. These CPU intensive functions are offloaded from the processor freeing up the CPUs resources to manage other more critical tasks. Intel also improved the networking by integrating some of the processors, denoted by an “F” at the end of the part number, to support Intel Omnipath or 100GB networking.
Overall, the Xeon Processor Scalable family has been optimized and designed for virtual environments and can be scaled with up to 3.4 times as many virtual machines as older servers. The performance gains are not just for virtualized data centers but for all levels of customers who will see drastic leaps in performance and capabilities across a wide range of workloads.