Scott Wasson, the highly respected technology journalist of The Tech Report, said he believes that the newly released Intel(NASDAQ:INTC)Xeon D processor family is the "future of the Xeon lineup." What does he mean by this? Wasson clarifies that the product is "poised to ship in the largest numbers overall once the market comes to understand it."
I am going to walk through some of his argument and also offer up some of my own thoughts on the subject.
Power efficiency galore
In the server market, the "total cost of ownership" equation is generally believed to be the biggest driver of purchase decisions. Total cost of ownership includes not just the upfront hardware acquisition cost but also -- according to Intel, citing an analysis from Amazon -- networking, power distribution, cooling, and power costs.
This is why Intel, as well as its competitors, try to focus on delivering not just more performance but more performance in an energy efficient manner.
According to Wasson, larger processor cores – like the latest Intel Broadwell CPU core – are "most power-efficient when operating at the lower end of their possible voltage and frequency ranges." He also goes on to note that Intel manufacturing technology advancements favor lower voltage operation.
Putting this into the context of Xeon D, Intel includes eight CPU cores that run at a base clock of 2GHz and can turbo up to 2.5GHz on all cores and 2.6GHz on a single core. Wasson points out that traditional dual-socket server chips can run at higher frequencies but that "those speeds come at less efficient operating points."
The Xeon D, unlike the more mainstream Xeon E5 processors, is limited to one processor per motherboard, while the Xeon E5 comes in dual and even quad processor configurations. Multi-processor configurations, Wasson points out, "tend to burn lots of power on their socket-to-socket interconnects."
What is Xeon D good for? High growth segments!
According to the following slide from the Intel investor meeting, the two largest markets for Intel server CPUs in 2014 were enterprise and technical computing. Cloud revenues were also substantial but a bit lower than technical computing and much lower than enterprise revenue.
That said, Intel does expect cloud revenues -- as well as technical computing -- to grow substantially through 2018. Enterprise revenue, which should largely consist of Xeon E5 and Xeon E7 chips, are expected to continue to grow, albeit at a much slower pace than the other segments.
If the Xeon D family turns out to be the "ultimate" cloud-server oriented server chip family, then I could certainly see it making up a significant chunk of Intel server processor revenue.
Another big advantage for Xeon D
If you follow the various Intel product lines, you will note that its multi-socket server processors often lag its PC/mobile processors significantly when it comes to adopting new manufacturing technologies and CPU architectures. For example, the 14-nanometer Xeon E5 family is expected to launch in early 2016, according to some leaked roadmaps.
In contrast, the first 14-nanometer Intel tablet/PC processors, featuring the new Broadwell core, showed up on shelves in late 2014 with a substantial rollout happening in early 2015.
The Xeon D adopted Intel 14-nanometer technology and the Broadwell CPU core well before the Xeon E5 and E7 are expected to, and not too long after PCs got both. I suspect that going forward, Xeon D will have an additional efficiency edge over the E5/E7 chips, because Intel will move these chips to newer processes and CPU architectures sooner.
Yet another big advantage
Another thing worth noting is that the Xeon D integrates more functionality onto a single chip than the Xeon E5/E7 chips do. For example, the current Xeon D chips feature an integrated 10 gigabit Ethernet controller, and variants coming later this year are said to include integrated cryptography accelerators.
I believe that Intel will bring out more variants of Xeon D with different types of accelerators integrated onto the silicon to target particular workloads. Intel is unlikely to play "mix and match" with various accelerator blocks on its Xeon E5/E7 chips, which tend to be straight-up CPUs.
Wasson's claim seems plausible
I find it hard to disagree with the idea that Xeon D will make up a significant portion of Intel's server CPU mix. I also think that Xeon E5/E7 will become a proportionally smaller part of Intel CPU shipments over the coming years.
Whether the various Xeon D processors become Intel's highest volume server chips is not yet clear. However, if Intel goes all-in on Xeon D and offers a broader range of core counts, integration options, and so on to its customers, Wasson should ultimately be proven correct.