What’s next for laptop CPUs?
We could see some surprising developments in the near future when it comes to mobile processors.
“The real movement doesn’t come until Cannonlake in late 2017. That’s when Intel will move to its 10nm process, improving performance-per-watt significantly.”
Intel
Intel’s processor strategy has, for the last few processor generations, relied on a ‘tick-tock’ strategy. An architecture update would be followed by a manufacturing process shrink, which would, in turn, be followed by another architecture update and so on. That’s set to change, as delays in Intel’s upgrade to a 10nm process have scuttled its original plans.
Right now, Intel’s top-of-the-line mobile processors (including the processors used in all the laptops reviewed) use its Skylake architecture, which, like its predecessor Broadwell, is manufactured with a 14nm process. It offered a number of modest improvements over the previous platform, including a significantly better GPU (Intel Graphics 530) and better power control to lower overall consumption.
Another upgrade is due towards the end of this year. The upcoming 14nm Kaby Lake processors will have integrated USB 3.1 support (other platforms require a third-party chip), another upgrade to the GPU with support for hardware decoding of 10-bit HEVC video as well as native support for Intel’s new Optane memory and flash storage architecture. In short, it’s another modest upgrade that will probably go unnoticed by most users (unless Optane really takes off, which, if Intel’s claims about its performance are in any way accurate, it might).
The real movement doesn’t come until Cannonlake in late 2017. That’s when Intel will move to its 10nm process, improving performance-per-watt significantly. We don’t know the clock speeds it will be available at, but you can expect a significant jump in performance.
Following Cannonlake, Intel has Icelake and Tigerlake in 2018 and 2019, respectively. These are architecture upgrades, although we don’t yet know what they include.
Intel even has a preliminary roadmap beyond that. Leaked reports indicate it plans to move to 7nm by 2020 – a mark that was once thought impossible on silicon. Manufacturing rival TSMC plans to go 7nm even earlier – as soon as next year, and into 5nm territory by 2020 (possibly using indium gallium arsenide in lieu of silicon). Beyond 5nm is even possible, with carbon nanotubes promising feature sizes as little as 1nm across. All of which is to say: Moore’s law has some juice in it yet, and Intel will likely be right there.
AMD
Although it hasn’t been a big player in the mobile space for some time, AMD still has the potential to make a comeback. For one, Intel’s substantial manufacturing process advantage over AMD is potentially set to disappear. This year, AMD is moving to a 14nm process, which is the same as Intel is using in its high-end processors. Current AMD CPUs use 32nm and 28nm processes.
With Samsung and TSMC starting 10nm manufacturing late this year and early next, AMD may even steal a jump of Intel, which won’t be making 10nm chips until mid-2017. Now that it’s no longer vested in GlobalFoundries (which has historically struggled to keep up with Intel), AMD can use these other manufacturers to potentially put out chips that rival or exceed the performance-per-watt of Intel.
In addition to the improved manufacturing process, AMD is moving to a completely new architecture. At the end of this year, AMD will introduce the new ‘Zen’ architecture, which will completely replace the current ‘Bulldozer’ core.
According to AMD, it will have up to 40% better performance-per-clock than the current gen, and will support simultaneous multithreading, similar to Intel’s hyperthreading. It will also support DDR4 and high-bandwidth memory (HBM). It will come on desktops first, with mobiles likely to follow next year.
With AMD already having a significant integrated GPU advantage over Intel, it will certainly be worth keeping an eye on its processors over the next few years.