New Xeon E5 v3 promises a respite for capacity planners

Intel Developer Forum is a major geekfest for the people interested in the guts of today’s computing infrastructure, and I want to call out the announcement there of the new Xeon E5.

Share

Intel Developer Forum is a major geekfest for the people interested in the guts of today’s computing infrastructure, and I want to call out the announcement there of the new Xeon E5.

While I’ve discussed it in more depth in an accompanying Quick Take just published on our main website, I wanted to add some additional comments on its implications for data centre operations, particularly in the areas of capacity planning and long-term capital budgeting.

For many years, each successive iteration of Intel’s and partners’ roadmaps has been quietly delivering a major benefit that seldom gets top billing – additional capacity within the same power and physical footprint, and the resulting ability for users from small enterprises to mega-scale service providers, to defer additional data spending capital expense.

In the case of the new E5 v3, the enabling technologies come together in a nearly perfect storm of improved performance coupled with actual lower system power consumption to make systems based on the new CPU particularly attractive to data centre management-facing demands from their business stakeholders for systems that cannot, for a variety of reasons, send to the cloud.

Based on a quick skim of the flood of third-party benchmarks beginning to appear, systems based on this new CPU and its accompanying platform elements, such as DDR 4 memory, are delivering performance of 25 – 40% over the previous generation E5 v2 along with lower system power consumption under load and lower idle power.

Since servers typically have a three- or four-year lifecycle, especially if they are leased, the actual impacts on mainstream data centres will be even more profound, as a large number of Westmere-based (32 nm server technology from a couple of generations ago) servers will come off of lease over the next year, and the throughput per watt performance for the new E5 v3 is probably conservatively a factor of at least 2X compared to this older base.

The net impact is that a capacity planner faced with escalating workloads has been effectively handed a “Get Out of Jail”[i] card for their next 30 – 100% of increased throughput from their current facilities. Good news and good economics for all concerned.

Posted by Richard Fichera

"Recommended For You"

Total boosts Pangea supercomputer with 4.4 petaflops of compute power Intel pushes 50-core server chips for exascale HPC programme