Per Eurocom, Intel is planning to launch a gaming chip with eight cores in the second half of 2018. I had originally assumed that this eight-core part was based on the company's upcoming Ice Lake architecture and manufactured using the company's 10nm+ technology.
However, BenchLife.info appears to have come across some information that indicates that the eight-core chip referenced by Eurocom isn't based on the Ice Lake architecture, but instead on the same Coffee Lake architecture that underlies Intel's upcoming six-core processors.
A sensible option
Intel's upcoming Coffee Lake architecture chips are expected to be manufactured in the company's third-generation 14-nanometer technology, known as 14nm++.
14nm++, per Intel, offers either 26% better transistor performance than its original 14-nanometer technology introduced back in 2014 at similar power consumption, or 52% lower power consumption at the same performance level.
In fact, 14nm++ is so good, at least according to Intel, that in terms of transistor performance, it's better than both Intel's first-generation 10-nanometer technology and its second-generation 10-nanometer technology, known as 10nm+.
From a cost perspective, Intel says that the cost per transistor of its newer 10-nanometer technology should be lower than that of its 14-nanometer technology and performance-enhanced derivatives. However, those claims are normalized for manufacturing yields.
The reality is that Intel's 14-nanometer technology has been in mass production since mid-2014, and by the time this rumored eight-core Coffee Lake chip should launch, the technology should be extremely mature. In contrast, Intel's issues in ramping its 10-nanometer technology are well-known, and low yield rates on the technology probably mean that the die cost of an eight-core Ice Lake chip built with 10nm+ is higher than the die cost of the rumored eight-core Coffee Lake chip using 14nm++.
Moreover, yield rates don't just affect product cost -- they can affect supply, too. Lower yield rates mean that for a given amount of silicon wafer production capacity, fewer salable processors can be produced. Such supply issues can negatively impact a company's ability to meet demand for its products, hurting its financial results and souring relationships with customers.
A company can theoretically just put much more capacity into place to try to compensate for low yield rates, but as yield rates improve, such a strategy would ultimately lead to excess manufacturing capacity -- something that can very negatively impact company profits.
So, if Intel's 10-nanometer technology isn't in great shape, Intel's management may have decided to take a more pragmatic approach by just building an improved, relatively low-risk product using its mature 14-nanometer technology rather than try to push 10-nanometer/10nm+ products out the door when those products simply can't be manufactured at reasonable costs or in the desired quantities.