Bloomberg recently published a nice piece that goes into some detail about how microprocessor giant Intel (NASDAQ:INTC) builds its chips. Although the article itself is informative in that it provides some insight into how extremely complex processors are made, there was one comment from an Intel employee that may have shed some light on a future Intel server processor.
The "mystery" 10 billion transistor chip
Bloomberg published a quote from Intel product manager Maria Lines in which she says the following:
The product that I was on several generations ago was about 2 billion transistors, and now the product I'm on today has 10 billion transistors.
A chip is made up of billions of individual elements known as transistors; a greater number of transistors typically means that a chip company has packed more features and functionality into a processor.
At any rate, Intel's recently released 24-core Broadwell-EX chip, which is manufactured in the company's 14-nanometer process technology, packs about 7.2 billion transistors in an area of about 456 square millimeters.
Intel's next-generation high-end server chip, which will be known as Skylake-EX and will also be manufactured in the company's 14-nanometer process technology, is expected to pack up to 28 cores on a single piece of silicon.
If we assume linear scaling of die size with core count, Skylake-EX should weigh in at around 8.4 billion transistors. Do note, however, that the cores that power Skylake-EX are expected to be more powerful than those inside of Broadwell-EX, and it is expected to feature beefed-up "non-core" parts as well.
I could see the chip easily packing 10 billion transistors.
I believe that the chip that Lines is referring to is, indeed, the next-generation Skylake-EX that's slated to launch during the second quarter of 2017.
How about a die size?
Assuming that Intel doesn't see a dramatic change in transistor density (in other words, chip area scales with transistor count), then the largest Skylake-EX chip should measure around 633 square millimeters in size -- a large boost from Broadwell-EX.
Generally speaking, Intel doesn't max out the transistor counts on the first generation of product built on a given manufacturing technology. This is for a couple of reasons.
The first is that by not maxing out the capabilities of a manufacturing technology on the first try, the company has headroom to deliver performance/power/feature improvements in a follow-on generation. The second is that manufacturing yields typically get better over time, so it becomes more economical to build larger chips later on a given manufacturing technology.
Looking to the future
With Skylake-EX, Intel will have virtually maxed out what it can do with its 14-nanometer technology. The next step for the company will be to migrate to its next-generation 10-nanometer technology, which should provide both a transistor-level performance boost as well as a significant area reduction.
The upshot is that Intel should be able to dramatically increase the core counts in its 10-nanometer server processors. This will ultimately mean more performance, further incentivizing server operators to upgrade to the company's latest processors.
Ashraf Eassa owns shares of Intel. The Motley Fool recommends Intel. Try any of our Foolish newsletter services free for 30 days. We Fools may not all hold the same opinions, but we all believe that considering a diverse range of insights makes us better investors. The Motley Fool has a disclosure policy.