How Intel’s 9th Gen chips show the way forward after Moore’s Law
Intel’s 9th Gen Core processors are here, and, as expected, they’re simply a refresh of the last generation of chips, with the same 14nm process the company’s been using since 2014.
Performance-wise, Intel says the new chips are squeezing out higher speeds; they’re supposedly up to 10 percent faster than last year’s models. There are also some nice improvements on the more granular side of things for particularly serious users, including the switch back to a solder thermal interface material (STIM) over paste.
But for the most part, the new chips have the same things last year’s chips had: more cores. And the reason is pretty simple: Intel still hasn’t managed to move on from its 14nm manufacturing node to the next step, its repeatedly delayed 10nm process.
Each processor manufacturing generation is determined by the manufacturing node, which describes the size of the minimum feature on the wafer of silicon. Generally, the smaller the node, the more technology can fit onto a chip, and the better the overall performance.
Intel is now on its third generation of products using this old architecture: the 14nm node dates back to 2014’s Broadwell processors, which means that Intel hasn’t advanced to a significantly smaller transistor size in almost half a decade. It’s something that stands in stark opposition to the long-vaunted Moore’s Law — Intel’s guiding star for the last 50 years — that won’t continue to hold for the future.
The added benefits of increased transistor count that come from shrinking die sizes — like decreased power demands, lower heat, improved yields, and cheaper cost per chip — will be harder to come by going forward. Instead of the classic tick-tock structure, where Intel would go down a step in architecture size for one generation and optimize it the next, we’re spending more and more time on the current step, and Intel has to continue to find new ways to step up performance. This year’s 9th Gen chips are technically a refreshed version of the 2017 Coffee Lake architecture, which Intel calls the 14nm++ node, which is the second major update of the original 14nm node.
None of this is new, exactly. We’ve known for a while that sooner or later, Moore’s Law would break down. You can physically make a transistor only so small, and it appears that while Intel hasn’t quite hit that yet, it’s certainly having trouble making the same kind of jumps it once did.
This means both the company and the customers are going to have to make adjustments about what to expect from future generations of chips as Intel is forced to look down other roads to improve performance and add new features to encourage customers to upgrade.
The new multicore strategy is only the beginning of that, as are other strategies like the focus on gigabit Wi-Fi that Intel highlighted with its latest laptop chips earlier this year or the unlikely partnership with AMD to include an onboard GPU for its more powerful mobile processors.
So far, Intel’s been able to make the stalled Core generations work with other improvements, like the multicore options mentioned above. That’s something we’re still seeing with the 9th Gen chips; the top i7 and i9 models now offer eight cores each. And that’s not even counting the Core X line (which Intel also just refreshed) that offers even more multicore power for demanding users.
Whether that’s sustainable in the long term, we’ll have to wait and see. But if this year’s batch is anything to go by, it seems like Intel’s future may be one of more gradual change, instead of a sprint toward the future.
Correction: Referred to “process” incorrectly as “die size”.