by Austin Hipes, Director of Field Engineering for NEI.
This revised excerpt was taken from an article that originally appeared in Virtualization Journal on Nov. 18, 2012.
The E5 family of Intel Xeon processors, which are based on Sandy Bridge architecture, is responsible for changing the way in which data-intensive applications run on AdvancedTCA hardware systems. Innovative networking is possible through 40-gigabit Ethernet, and its features allow for advanced cloud solutions using virtualization methods. By deploying virtual machines in a private cloud environment, Sandy Bridge processors installed on ATCA platforms create the perfect infrastructure to apply these solutions.
The Xeon E5-2600 series CPUs house up to 8 cores, each running over 50 percent faster than their Xeon 5600 predecessors. As a result, much better server performance is delivered to an enterprise-class marketplace. In addition, the new enterprise server platforms support dual in-line memory modules (DIMMs) up to 32 GB, increasing memory capacity from 288 GB to 768 GB using 24 slots. With more limited board real estate, E5-based AdvancedTCA compute blades are expected to support 256 GB in 16 VLP RDIMM slots when the processors launch, representing an increase of 40 percent over existing blades today.
Power efficiency gains are another benefit of the new Sandy Bridge microarchitecture. There is up to a 70 percent performance gain per watt provided by the E5 family over previous generations of CPUs. It is possible for communications OEMs to supply service providers with power efficient dual processor blades that completely meet or beat AdvancedTCA power specifications because of this advancement.
The main advantage of Sandy Bridge, however, is in the E5-2600’s integrated I/O, which lets designers reduce latency and increase bandwidth significantly. The 40G fabric of ATCA has been backplane-ready since 2010 in anticipation of the release of updated PICMG specifications, and during that time solution providers have been looking for ways to eliminate bottlenecks to take advantage of as much of that fabric as possible. Now, PCI-Express 3.0 has been integrated into each Xeon processor with 40 lanes on board, and QuickPath Interconnects (QPIs) have been added to link each CPU together. The result is a reduction in I/O bottlenecks, increased throughput speeds, and an I/O latency reduction of up to 30 percent. The standard Xeon E5-2600 dual CPU configuration has up to 80 lanes of PCIe Gen3 available, providing more than a 200 percent increase in throughput over previous generations of architecture.
The overall result of the new Sandy Bridge E5-2600 series design is a much higher I/O throughput. Now, ATCA blades will be able to deliver over 10 Gb/s per node, a critical milestone for wireless video applications that service providers have been very anxious to roll out. Improvements in overall performance and better power efficiencies are significant achievements in-and-of themselves, but being able to match processor capacity to I/O capacity makes for far greater gains in application throughput.
For more information, Contact NEI online or by calling (877) 792-9099.