Intel Announces Xeon E7 v4 Processors For Mission-Critical Computing And The Cloud
Intel Announces Xeon E7 v4 Processors
Though they are a different family of products, the Xeon Processor E7 v4 family still leverages Intel’s Broadwell core technology like the E5 series, specifically, the Broadwell-EX core. Since we’ve covered Broadwell many times in the past, we rehash that here again. If you’d like some additional background on Broadwell, however, we’d suggest reading our launch coverage for both the mobile and desktop parts, and our previous Xeon coverage. Our launch coverage of last year's E7 v3 series Xeon covers many of the platform details as well, should you want a refresher.
The Xeon Processor E7 v4 family is designed for mission-critical server applications, big data analytics, and the cloud. The processors are meant to scale-up data centers and provide maximum memory capacities and the fastest transaction times, in order to speed real-time insight and analyze massive data sets in industries like banking, transportation, telecom, and health care, among others.
Servers with Intel Xeon Processor E7 v4 series processors can support up to 24TB of memory (3TB per socket), which is double the previous generation. They support up to 8 sockets in a single system and are drop-in compatible with existing Brickland-based platforms, after a BIOS / firmware update. In most segments, Xeon Processor E7 v4 series processors typically feature more cores and cache than their previous-generation counterparts, and occasionally higher clocks as well. Xeon E7 v4 series processors with up to 24 physical cores (48 threads) and 60MB of cache will be available.
Previous-gen Haswell-based Xeons supported a 2S Cluster-on-Die mode, designed for more efficient use of the processor's LLC, or last level cache. With the new E7 v4 series, however, a 4S Cluster-on-Die mode is now supported. In a 4S cluster, each cluster operates as an independent caching agent, and the OS creates associated domains, so that most cache accesses remain within the designated cluster. As a results, memory subsystem and LLC access latencies are reduced, because cache slices are localized to the cores and the number of threads seen by each home agent is reduced, which increases the chance that memory requests hit open pages in the memory controller.
In addition to the performance enhancements that come from the additional cores and cache in the processors, and the Broadwell architecture itself, Intel continues to tune its tools to optimize performance on the many-core Xeon Processor E7 v4 series as well. The Intel VTune Amplifier for 2016, for example, can be used to analyze workloads and identify areas that can be tuned or re-ordered to better utilize the processors, and ultimately improve efficiency and overall performance.