On May 30, microprocessor giant Intel (NASDAQ:INTC) announced its Core X series of processors and the accompanying X299 platform for high-end desktop personal computers.
The lineup consists of two distinct families of processors. The first is Skylake-X, which comes in configurations with as few as six processor cores and as many as 18 processor cores (though the 18-core models are slated to become available later).
The second is Kaby Lake-X, which includes a four-core/four-thread chip in the form of the Core i5-7640X and a four-core/eight-thread chip in the form of the Core i7-7740X.
The Kaby Lake-X chips are quite controversial among many in the technology community because they are merely the same Core i5-7600 and Core i7-7700K chips the launched back in January, respectively, but repackaged to work in the more expensive X299 motherboards, rather than the lower-cost Z270 motherboards.
Here's why I think Intel chose to release these chips.
The entry-level X299 parts
Intel clearly wanted to have "entry level" parts for its high-end desktop platform, and there are certainly good reasons to want them.
For example, an individual building a computer around the X299 platform might want a computer now, but may not have enough money for the higher core count part that they may eventually want. So, a reasonable option for that customer would be to build the system with a relatively low-cost "stop gap" part and then, at some point down the line, upgrade to something better.
Alternatively, customers might simply want the features/capabilities that the Core i5-7600K/Core i7-7700K offer, but would like to pair them with more feature-packed motherboards. Buying a 7640X or a 7740X and pairing it with an X299 motherboard (high-end desktop motherboards tend to be more feature-rich on average than mainstream desktop motherboards) would be a good solution for such a potential buyer.
Now, you might be thinking that it might make more sense for Intel to simply offer quad-core versions of the Skylake-X chips than to do the work to bring the mainstream Kaby Lake-S chips over to the X299 platform. But it doesn't.
Two big reasons
The first reason that Intel likely didn't simply offer a quad-core Skylake-X part as the entry-level chip for X299 is economics.
The Skylake-X parts are based on two silicon die: the first is Intel's LCC chip, which has 10 cores, and the second is the company's HCC chip, which has 18 cores on the die. Per InstLatX64 on social media, the chip size of the LCC Skylake-X is north of 300 square millimeters.
Intel would have to disable a substantial portion of this large, expensive-to-manufacture chip if it wanted to use it to offer quad-core parts on the X299 platform.
That's not good for Intel's product margins, particularly as it's unlikely that its manufacturing yields on its 14-nanometer+ technology are bad enough that it produces many parts with six defective cores.
By contrast, Intel's quad-core Kaby Lake die measures in at around 126 square millimeters in the same 14-nanometer+ technology that the Skylake-X parts are manufactured in. Now, much of that die area is taken up by a graphics processor that's disabled, but even with that part of the chip turned off, the chip is still small enough to begin with, and sells for a hefty enough sum, that Kaby Lake-X represents a financial "win" for Intel.
And one more thing
A final reason that the Kaby Lake-X processors make more sense than a heavily disabled Skylake-X part is Intel's segmentation strategy. The Kaby Lake-X chips only have 16 PCI Express lanes on the silicon die, while the lowest-end six-core Skylake-X has 28 of its 44 (usable) lanes enabled.
So, Intel's sales pitch to customers, then, is that they can have 50% more cores and 12 more PCI Express lanes for about $40-$50 more than the top Kaby Lake-X part. This, Intel probably hopes, will encourage upsell to the six-core part, ultimately enriching its product mix.