If you recall SemiAccurate’s exclusive news about Cannonlake’s triumphant return from the fab, for some values of triumphant, we were a tad sceptical. That was 355 days ago though and our opinion hasn’t gotten better since then. We also told you about how Coffee Lake, the second 14nm stopgap product, was going to hit Intel in the margins.
Last spring Intel went into full damage control mode with their Manufacturing Day inventing the term Hyperscaling to cover up for massive process problems. Officially Hyperscaling means nothing is wrong with Moores Law and more importantly the financial benefits it brings. If you actually look at the numbers however, they contradict Intel’s financial promises and worse yet contradict 25+ years of statements the speakers themselves gave. If you are overly skeptical you might claim Hyperscaling was just made to keep financial analysts from understanding what was really going on.
One thing that was quite clear is that Intel’s 10nm process was having severe problems. The year long extra gestation of the 14nm process, promised in late 2013 but delivered in late 2014, has become 2+ years for 10nm. Cannonlake, the 10nm shrink of Skylake, which launched in August of 2015, was meant to be an early 2016 product. Broadwell’s delay pushed Skylake out which in turn pushed Cannon out a little. The next two years of 10nm delay is not intentional.
Nor was Kaby Lake, Coffee Lake, and now Whiskey Lake. Officially Cannon Lake is a late 2018 product now, something SemiAccurate has no faith in Intel actually pulling off. Why? Our sources still say the 10nm process is “broken”, something the only major Intel Foundry customer reiterated before they jumped ship. Intel has not officially acknowledged this relationship’s end yet, just like the last two.
So why is 10nm so broken? One comment by an knowledgeable insider gave us the clue we needed.
Note: The following is for professional and student level subscribers.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- Intel kills off the 10nm process - Oct 22, 2018
- HyperX ships it’s 60 millionth enthusiast memory module - Oct 15, 2018
- Bittware/Nallatech water cools 300W of Xilinx FPGA - Oct 12, 2018
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018