Intel Xeon Processor E5 v4 Family Debut: Dual E5-2697 v4 With 72 Threads Tested

While being briefed on the new Intel Xeon E5 v4 processor family, we also had the chance to get some hands-on time with systems equipped with the processors and were shown a handful of demos. The demos consisted of various use-cases designed to illustrate the performance potential of the processors and platform.



There were a number of 2U servers on hand, powered by Xeon E5 v4 series processors and packed with various amounts of RAM and storage. The servers themselves were non-descript, but we thought you’d like to take a look at them nonetheless.


If you haven’t experienced a server of this type first hand, know that they are densely packed (they’re only 88mm high and 487mm wide), but many feature multiple processors. The machines pictured here were outfitted with dual Xeon E5 2699 v4 processors, each with 22 physical cores, for a grand total of 44 cores (88 threads). They also feature 256GB of Hynix DDR4 RAM (4 sticks per CPU) and all solid state storage, connected through an Intel Grantley-EP based motherboard.


A system with all of those resources at its disposal will have some serious horsepower. If you’re a desktop enthusiast and would like a chuckle, here’s a look at CPU resources in Windows Task Manager in a system that supports 88 threads.

And here's was a 2P E5-2697 v4 base system can do in Cinebench -- this is wild to see in action.

Intel Xeon E5 v4 36-Core, 72-Thread 2U Server Slays Cinebench

The performance and benchmark tests being shown in the demos included Linpack, a financial analysis test, an iSCSI random read test (using a massive array of SSDs and trio of 40GbE links), and a dual-port NVMe demo (Intel is also launching some new enterprise-class solid state drives today).




In the financial analysis demo, the Xeon E5 v4 series system was up to 46% faster than the E5 v3-based system, due to architectural enhancements and the increases memory bandwidth / transaction speed on the newer Xeon E5 v4 platform. Linpack peaked in the 437 GFlops range, and the iSCSI test (using the Intel Storage Performance Development Kit, or SPDK), the system sustained over 3.2M IOPs with only 42% CPU utilization (15 cores being utilized).