Earlier this month, chip giant Intel (NASDAQ:INTC) announced a new set of processors for the low-cost personal computer market, known as Gemini Lake and marketed under the company's Pentium Silver and Celeron brands.
The chips include some interesting improvements in both feature integration and CPU performance. Based on some early performance tests, the Gemini Lake chips look like reasonable generational improvements and should serve the needs of low-cost PCs well enough.
What is particularly interesting about Gemini Lake, though, is that it is reportedly manufactured using Intel's 14-nanometer technology -- that is, the same 14-nanometer technology that's been in manufacturing since 2015 -- rather than the company's higher-performance 14-nanometer+ or 14-nanometer++ technologies that are used to build the company's higher-end Core PC processors and its Xeon processors.
Had Intel built Gemini Lake using 14-nanometer+ or 14-nanometer++, it's likely that the chip would've delivered better performance and power efficiency.
So, why didn't Intel do that?
Cost is king
The thing to remember is that Gemini Lake, as well as its predecessors and, of course, its successors, is aimed at very cost-sensitive computers. Intel needs to deliver improved features and capabilities, but the chips need to fit into tight cost structures.
Indeed, computers powered by Gemini Lake's predecessor, Apollo Lake, can be found for around $300. These machines aren't the best things around -- they're often bulky, have slow storage, poor quality displays, and generally cheap build quality, but there is certainly a market for them, particularly in developing countries where $300 represents a significant portion of somebody's monthly paycheck.
Since Intel's 14-nanometer+ and 14-nanometer++ added new features and technologies to boost performance, I wouldn't be surprised to learn that those performance enhancements also led to increased manufacturing complexity and, therefore, chip manufacturing cost.
The cost increases likely aren't huge in going from 14-nanometer to 14-nanometer+, but if such a move adds even, say, a dollar or two to the overall chip manufacturing cost without meaningfully changing its competitive positioning in the marketplace, then Intel would just be leaving profit on the table.
What comes next?
Per a tweet from FanlessTech, an individual who has posted reliable information about future Intel products in the past, the successor to Gemini Lake is known as Gemini Lake+, and then the part after that is known as Jasper Lake.
I think that Gemini Lake+ could be a good candidate for Intel's 14-nanometer+ technology, particularly if Intel ultimately reuses the same basic Gemini Lake architecture but still wants to deliver a reasonable generational performance boost (14-nanometer+, per Intel, is about 12% faster than the original 14-nanometer technology).
Jasper Lake, however, could very well be a part built using the company's 10-nanometer technology. Since a recently leaked Intel road map shows Gemini Lake holding the line throughout the entirety of 2018, Gemini Lake+ is probably going to be the 2019 part, and then Jasper Lake the part that lasts throughout 2020.
Although Intel's 10-nanometer technology still seems to need a lot more work before it's cost effective, I think the odds are reasonable that it will be ready to use (that is, cost effective enough) for a Pentium Silver/Celeron product that launches in the 2020 time frame.