Qualcomm will transition to 10nm for products entering the market next year. However 7480 will still be on TSMC 28nm! How does this translate into cost per unit?
2 replies (most recent on top)
Cost-per-gate actually started to rise after 28nm. 20nm was such a dead loss that many skipped it and went to 16/14, only to find that while leakage power went down, dynamic went up.
And now at 28nm, companies like TSMC are offering ULP solutions that have attractive power, or GF 20nm FDSOI which is getting much traction for IOT (IOT spends most of it's time "off" or in retention-mode, so leakage and power island management becomes more important that dynamic until you hit a max desired power limit for the battery system)
28nm may actually be the sweet spot in terms of cost per gate for the next decade:
http://www.semi.org/en/node/57416
But none of this helps Intel who aren't exactly known for their "low power" capability or the empowering of RTL engineers when a new interchangeable army of H1B drones can be dropped into the machine every 3-to-6 years.
Newer mode is a lot more expensive but with TSMC yields the much higher DPW translates into huge profits for Qualcomm! Intel 28nm is like trying to sell 32nm x86 to Apple for MacBook. You give it away at a loss, LOL