We delve into Intel’s latest manufacturing process to find out what’s next for your PCs and laptops – and to explore the company’s plans for smartphones and tablets.
There are only a handful of firms that move the semiconductor industry forward and, when it comes to mainstream silicon, only one company is pushing the envelope right now – Intel.
The Californian chipmaker has recently unleashed its latest big development: a move to 14nm manufacturing. It’s a significant step forward from the 22nm process that underpins the Ivy Bridge and Haswell chips that have been found inside desktops and laptops over the past couple of years.
This big development will improve Intel’s CPU performance significantly – but how does this new technology work?
What is 14nm?
The term “14nm” refers to the size of the transistors – the building blocks of a processor – and Intel’s latest size reduction is governed by Moore’s Law, which states that the number of transistors in an integrated circuit will double every two years. Make no mistake about the tiny size of each transistor: a 14nm unit is hundreds of times smaller than the average human hair, which is about 60,000nm in diameter.
Transistors control the electrical signals that flow through processors, acting as switches that turn on or switch off the flow of electricity in order to interpret instructions that arrive at the processor from other areas of the PC. Transistors handle impossibly tiny bits of work, with 1.3 billion of the units used to build a typical Core M processor.
Intel’s move to a 14nm process has seen the individual components inside each transistor reduce in size: the size of the fins, gates and interconnects have reduced by between 22% and 35%, and there are now two fins per transistor rather than three – another step to improve density and efficiency.
Shrinking these individual parts is important, as it reduces the amount of electricity required for those fins, gates and interconnectors to function, which means Intel can achieve the same performance level as the last generation’s components with smaller transistors and less juice.
That, in turn, means that even more transistors can be installed into the same surface area – so that same area can complete more instructions while proving more power efficient – and, hopefully, fulfilling the prophecies laid down by Moore’s Law.
It also helps that Intel’s smaller transistors leak less electricity, which helps reduce inefficiencies and heat consumption across the silicon.
The smaller Intel goes, though, the harder it’s got to work to meet its targets – the firm admits that this has been its toughest challenge yet when it comes to process manufacturing.
The smaller wafers are trickier to manufacture successfully, which means that Intel doesn’t get as many usable parts out of every batch – in tech terms, that’s described as a bad yield.
The 14nm parts started life with much poorer yields than their 22nm predecessors, and they’re only just catching up – and won’t reach parity until the end of the year. It also doesn’t help that the preceding 22nm process was Intel’s most successful ever in terms of yields.
The poor yields mean that Intel has modified its original 14nm plans. Intel wanted to unveil more of its 14nm Broadwell parts in 2014, but it just can’t make enough of them successfully, so the decision was made to release Core M parts this year and delay the rest.
It’s a tricky, expensive road, but there’s a light at the end of the 14nm tunnel. The cost of entire wafers of silicon may be increasing, but Intel is doing more with less – so its cost per transistor continues to decline, and it’ll only improve as yields get higher.
A Big Move in Mobile – Core M
The Core M parts that Intel can manufacturer right now are about to appear in a host of hybrids and Ultrabooks. These processors are low-power parts designed for slim, light systems, and they deploy some of the key advantages of a 14nm process to improve performance across the board.
Core M has hugely reduced power consumption over the last generation thanks to the drop in transistor size and, despite that, Intel claims sizeable performance improvements too – along with a big leap in battery life thanks to the lower power consumption.
The first wave of Core M devices include sub-10mm hybrids from Lenovo – such as the gorgeous Yoga 3 Pro – and other devices from Asus, Dell and Acer, and they’re an exciting glimpse of Intel’s 14nm-powered future.
Broadwell and Beyond
Plenty of other Broadwell-based chips are coming down the pipe for 2015. Additional low-power chips for Ultrabooks and NUC PCs will arrive in the first quarter of 2015, and a further set of low-TDP parts designed for small-form-factor PCs will arrive in the second quarter of next year. Both of these releases use the BGA package, which means they’re soldered to motherboards, and both consume more power than Core M parts.
Traditional LGA 1150 desktop processors are also slated for the second quarter of 2015. They’ll top out at 65W, which is 19W less than Haswell equivalents.
Beyond Broadwell, expect Intel’s next big tock – Skylake. These chips will introduce a new architecture with changes to voltage regulation and support for PCI Express 4.0, Thunderbolt 3.0, SATA Express and improved integrated graphics on chips that retain the 14nm manufacturing process.
Broadwell’s delays mean that Skylake will arrive unusually quickly, with the first parts debuting by the end of 2015. After Skylake comes Cannonlake, which is a tock rather than a tick – so it’ll shrink the Skylake architecture to 10nm rather than introducing many new features.
An ARMs race
Intel might have forged ahead with its 14nm process, but its move towards greater efficiency and performance – and, therefore, towards mobile – brings it in line with ARM.
The Cambridge-based firm rules the roost in tablets and smartphones and has made designs towards larger systems, so it’s no surprise that it’ll cross swords with the new, mobile-friendly Intel.
ARM’s current lineup is dominated by its Cortex-A parts, which range from the modest A5 to the powerful A17 with five other chips in between. The current Cortexes are fine for a wide range of smartphones and tablets – Samsung’s smartphones and Nvidia’s Tegra parts both use them – but they’re 32-bit parts that can’t really compete with what Intel offers, as they’re all 32-bit chips with smaller caches and often with lower clock speeds.
ARM’s planned roadmap could change all that, and bring the firm into direct conflict with Intel’s forthcoming parts. Its Cortex-A53 and A57 chips will support 64-bit computing, which means that they can address larger amounts of memory as long as software is coded to support 64-bit processors.
ARM’s move to 64-bit processing means that smartphones and tablets will become more powerful, which could scupper Intel’s plans to further grow in portables. That’s not the only potential implication, though, as ARM’s 64-bit chips could easily make the move to modest laptops – an area where Intel has dominated with Celeron, Pentium and Core i3 processors.
And ARM already has a bit of form when it comes to areas that have traditionally been dominated by big players like Intel: AMD has designed an ARM-powered CPU for servers and, on launching its Operaton A1100 chip, it alluded to huge cost savings when compared to Intel. AMD reckons that chip provides great performance for applications that need low-power computing, huge memory support and plenty of storage.
What Happens Next?
Traditional processors and mobile hardware used to be almost entirely separate, but a host of factors mean these worlds are converging: the chips used inside PCs and laptops are being developed for power efficiency rather than performance, the hardware inside smartphones and tablets is becoming increasingly powerful, and the devices themselves are now blurring form factors – we’re seeing more hybrids and phablets than ever.
Given these conditions, it’s no surprise that the main firms are converging, with ARM blurring the lines between mobile processors and traditional chips, and Intel further shrinking its own components to improve performance and efficiency.
Intel’s move to 14nm is a big jump – an increasingly difficult move that they’ve pulled off with aplomb – and it’s going to pay dividends in all kinds of devices, from desktop systems and Ultrabooks to tablets and hybrids. It’s the clearest indication yet that Intel sees its future in mobile.
Next, read MacBook Pro vs MacBook Air