In the past, processor giant Intel (NASDAQ:INTC) developed what it referred to as a "converged processor core." The idea is that one single high-performance, power-efficient processor core could serve as the foundation for a whole family of products ranging from 2-in-1 convertible PCs all the way through servers.
Now, for each of these different segments, Intel has traditionally turned many knobs such as frequency, core count, and the rest of the "stuff" that goes into the chips aimed at the various market segments. For example, although the 18-core Xeon E5-2699 v3 and the dual core Core i7-4650U both use the company's "Haswell" CPU core, they're aimed at vastly different markets with very different performance requirements.
Interestingly enough, though, at a technical session hosted at Intel's recent developer forum, Julius Mandelblat, senior principal engineer with Intel, revealed that the company did things a bit differently this time around with Skylake.
Skylake becomes "configurable"
"Very soon we discovered that our desire to give our best solution for servers and our best solution for clients sometimes causes painful compromises," Mandelblat said during the presentation.
To get around such compromises, Mandelblat says that Intel decided to do what is known as a "configurable core." This, he notes, means that the company works on just one main processor core design but that it comes in two distinct versions: one targeted at traditional client workloads such as laptops, and the other targeted at servers.
Although Mandelblat didn't disclose the features specific to the server configuration of the Skylake processor (that will have to wait for when Intel launches the chips, probably sometime in 2017), he did indicate that the server configuration of Skylake has significantly greater capabilities than the version aimed at client workloads.
Here's why this makes sense
Intel's data-center business is, in my view, a great business. The division consistently delivers leadership products in a timely fashion, and it generates an immense amount of revenue and profit for the company.
In fact, last year, Intel's data-center group pulled in $14.39 billion in revenue and generated nearly $7.3 billion in operating income. If Intel's data-center group were a standalone entity, then -- excluding Intel in its entirety -- it would be the world's sixth largest semiconductor company (according to data from research from IC Insights).
It's clear that at this point, the revenue and profits that Intel has riding on its data-center group are so high that it only makes sense that this group gets a CPU core that doesn't feature the "painful compromises" Mandelblat referred to in his talk.
What does this mean for Intel's server business?
As an Intel investor, I'm glad to see that the company is willing to invest extra in making sure it has the best chip designs for its various market segments, widening and deepening its "moat," so to speak. The better tailored Intel's chips are for specific market segments, the better it would seem that the company is positioned to maintain high levels of market segment share.
Indeed, since Intel commands the lion's share of the revenue and, more importantly, the profit that is to be had in the server processor market, there will always be competition trying to take share from Intel.
That said, I'm not too worried. Intel's track record in the server market is excellent, it spends substantially on data center specific projects and technologies, and its strategy looks spot-on. Although as an Intel shareholder I'm going to pay attention to any potential competitive threats, I'm not too worried.