It begins with Core M, but it will eventually extend to encompass Core i3/i5/i7 CPUs for laptops and desktops, Xeon chips for workstations and servers, and probably eventually some low-end Celeron and Pentium parts as well. Like the current 22nm process, Intel is planning to use 14nm to make chips going all the way up and down its product stack. Improvements in 14nm: Sticking to Moore's Law
The Core M chips that are launching do a good job of showing what Broadwell can do that Haswell can't (and Intel tells us that PC OEMs are asking for them), but if Intel was in a position to replicate Haswell's relatively wide and smooth launch, it almost certainly would. That said, yield is not expected to "meet the needs of multiple 14nm product ramps" until the first half of 2015. It's Intel's way of saying "hey, 14nm started a little rough, but we'll keep chipping away at the problems until we've cleared them up." Intel points out that its 22nm process had the lowest defect density of any process Intel ever used, despite some troubles early on. Unsurprisingly, Intel is choosing to focus on the positive things here, and there's plenty about 14nm that looks pretty good for the company. Other chips waited until later (and some are still waiting), but they're the ones we've always had to wait for-budget Pentium and Celeron chips for desktops and laptops and Haswell-E/EP/EX Xeon chips for high-end workstations and servers. Intel announced 39 different Haswell chips across its quad-core and dual-core offerings at launch, ranging from low-voltage tablet and convertible parts all the way up to high-end desktop chips.
The bad news is that, whatever yields are actually at right now, they're not enough to support a launch lineup as large as the Haswell launch was last year. "Further improvements" to the process are still being made, and, at some point in 2015, Intel thinks yields will improve to roughly the level that the 22nm process is at right now. The good news is that Intel has 14nm production in what it deems a "healthy range," and the "lead product" for 14nm is in volume production and shipping to PC OEMs now-that would be the Core M, which we'll discuss in more detail later. That delay, which was supposed to move the planned start of mass production from the fourth quarter of 2013 to the first quarter of 2014, was due to a "defect density issue" that was more difficult to fix than Intel anticipated. Basically the only thing we've heard about it up until now is that it's the reason Broadwell is being held up.
The state of 14nm Know your codenamesīefore we talk about the new stuff Intel is doing in its 14nm process, we should talk about the state of that process. Intel isn’t giving out specific performance or power consumption numbers for its new processors just yet (expect some of that at the Intel Developer Forum early next month), but the company has provided many more details on what we can expect. The jump between 2012’s Ivy Bridge architecture and 2013’s Haswell architecture wasn’t huge, but for many Ultrabooks it made the difference between a mediocre product and a good one.īroadwell’s months-long delay has stalled this kind of progress (not that some haven’t tried to work around it anyway), but we’ll finally begin getting new chips later this year and into early 2015. The major industry players-everyone from Microsoft to Dell to Apple-depend on Intel’s progress to refine their own products. That's what makes Intel's progress important, and that's why we’re champing at the bit to get the Broadwell architecture and see Intel’s new 14nm manufacturing process in action. Its glory may be just a little faded these days, primarily because of a drooping PC market, tablet chips that are actually losing the company money, and a continuing smartphone boom that Intel's still scrambling to get a piece of, but something like 315 million PCs were sold worldwide in 2013, and the lion's share still has Intel inside. Usually our face-to-face interactions with tech company employees take place on neutral ground-an anonymous hotel room, convention hall, or Manhattan PR office-but two-and-change days on Intel's home turf really drove home the size of its operation. I saw several peppered throughout suburban Portland, and that's not even counting the gargantuan Intel-branded factory construction site I jogged by the next morning, or Intel's other facilities worldwide.
It was my first-ever visit to Intel's campus. Last week I flew from New Jersey to Portland, Oregon, to get briefed by Intel PR reps and engineers about the company's next-generation CPUs and the new manufacturing process behind them.