Is Intel’s Broadwell CPU family really badly delayed or are things what Intel claims they are, essentially according to plan? In short they are both, the chips are going in to production really soon but they won’t be out until Q4.
There has been a lot of rumors and FUD about Broadwell since Intel claimed that there would be roughly a 1Q delay to the start of production. You might recall Broadwell parts were supposed to start production, SemiAccurate assumes this to mean wafers in, late in 2013. Lets say December 1 for the sake of argument but a few weeks plus or minus makes no difference. With that start date, the first Broadwell parts were supposed to be trickling out in Q2 with more models following throughout the year although Intel had always stated a 2H/2014 release.
This means a roughly 6 month lag between wafers in and devices on the shelf for consumers. Three or so of these months are taken up by the semiconductor fabrication process, 10-12 weeks being the norm for 40nm and each successive node tends to add a little bit here and there. So that leaves about three months for Intel to stockpile enough CPUs to fulfill large OEM needs. This is once again a very normal thing that Intel has down to an art.
So if wafers in on December 1 means laptops on shelves on June 1, and subsequent models launching after that, we have a baseline. If you look back to how the 22nm Ivy Bridge launch happened, it was both a mess and slow. Wafers did go in during Q4 and laptops did start coming out in Q2, April 23 to be exact.
From there supply was low, models scarce, and over the next several quarters most Intel lines transitioned from Sandy to Ivy. A few more models trickled out in May while the volume 2C SKUs didn’t show up until September 3, 2012. For those counting that is about 9 months from wafers in to volume lines released but the initial trickle of parts only took about five months.
Now lets look at the state of Broadwell and the new 14nm process in light of all this. If all went swimmingly and wafers went in on the same WW (Work Week) as Ivy did, that would mean best case of a late April 2014 release. SemiAccurate has reason to believe that the 14nm wafers were due to start a bit later than 22nm did so lets call it mid-May. That was the plan but as Intel has previously admitted, the 14nm process had a bit of a “whoopsie” and is going to be delayed a quarter.
According to Brian Krzanich, Broadwell was supposed to start production later in the first quarter and, “still second half of 2014. Squarely on target.” So if wafers go in on March 30, six months is September 30 for the first devices on the shelf. That is Q3 but barely Q3 which is in 2H for those unclear on the numbering system. Any hiccup would mean Q4 and given some things beyond the scope of this article, Intel is sorely in need of a pat on the back and some water but that too is still in 2H.
But do note the word “still” in referencing the 2H/2014 release, it looks like Intel knew it would be tight on Q2 so they were planning on Q3 from the start. Why? 14nm process time didn’t increase by that much so that would mean the ramp is slow, something that SemiAccurate has been hearing from many sources. So far so good.
If that 5-6 months for 22nm Ivy launch stretches by just a little, late Q1 wafers in mean Q4 wafers out and the rest of the models trickling out in the next quarter or two. That would be Q1 and Q2 of 2015. While some people are running around like they just hit the lottery with their “discovery” that Broadwell would be Q4, if they had half a brain and understood the process they would realize there is no new news. In fact Intel’s CEO said exactly that in their Q4/2013 analyst call! When the CEO publicly says something over a month prior, if you have more than two brain cells to rub together it is hardly shocking that it comes about. That last bit however explains the furor over the, “delay” that all the news sites are hot and bothered about, that would be cranial capacity.
This is the long way of saying there is no delay. Intel said that it was a 2H product and it is a 2H product. It may have moved from the first day of Q3 to some time in Q4 but this is exactly what Intel said in their analyst call. One quarter delay for wafers in resulting in one quarter delay to products on sale is hardly the stuff of wonder and deep science.
Add to this another wrinkle of product timings. Laptops have one big cycle for the second half of the year and Broadwell is exclusively a laptop product.
That would be the so called “back to school” cycle that determines what will be on sale from August or so until roughly Chinese New Year. If a supplier can not get a part out in volume by June or possibly July, it is effectively dead until late Q4. A precious few models are upgraded in mid-Q4 in the hope of Christmas and New Year sales but most lines won’t be touched until late Q1.
If Broadwell was on time, Intel could ramp volume on a few SKUs to hit the back to school window. They missed badly so Q3 is effectively a dead zone even if they could launch then. If the next opportunity to sell a CPU is mid-December, and low volumes at that, why not take a little more time and increase yields? Even if you do perfect 14nm and can make volume, no one is going to buy them anyway.
So in light of all this, all we can say is, “deep breath people, chill”. There is a delay of Broadwell, it is process related, but it is exactly what Intel said it would be over a month ago. The process is delayed a quarter, Broadwell is still 2H/2014 even if that is a serious blurring of the lines, and nothing has changed. Anyone who thinks this is something new and shocking is simply ignorant of how things work in semiconductors.S|A
Have you signed up for our newsletter yet?
Did you know that you can access all our past subscription-only articles with a simple Student Membership for 100 USD per year? If you want in-depth analysis and exclusive exclusives, we don’t make the news, we just report it so there is no guarantee when exclusives are added to the Professional level but that’s where you’ll find the deep dive analysis.
Latest posts by Charlie Demerjian (see all)
- Thing go bump(gate) in the night for Nvidia’s GP100 Pascal GPU - May 3, 2016
- Cavium’s Octeon TX blends compute and packet moving - May 2, 2016
- Qualcomm releases Zeroth API to developers - May 2, 2016
- Another detail about Qualcomm server SoCs revealed - Apr 27, 2016
- AMD finally really honestly launches the dual Fury - Apr 26, 2016