Apple and Qualcomm are now testing 10nm pilot runs off of the Samsung process and the TSMC process in order to correctly award next year's production orders.
Rumor has it that low yields and overheating issues exist with at least one of these sets of submissions. Enough so to make it necessary to get a fresh large production sample from both companies, letting them fix everything they can in the meanwhile.A smart company (and Apple is a smart company, as is Qualcomm) will also be doing risk analysis on this new lithography level to see if it could possibly bite them in the ass like 20nm TSMC did for the Qualcomm 810 year before last with its rampant overheating (which was never totally fixed by TSMC, btw).
Apple, if not satisfied that 10nm is a good "smart" move from either vendor can certainly stop at 14nm level and simply wait until the non-silicone 7nm stuff is ready in 3 years. Or, both ARM and Apple can redesign their chipsets to a massively multicore arrangement, to help with the 10nm overheating issues.
NOTE: APPLE JUST ANNOUNCED A HEXACORE PRODUCT DESIGN FOR PHONES AND AN OCTACORE DESIGN FOR TABLETS/LAPTOPSThe monkey is now squarely on the backs of Samsung and TSMC to make their 10nm performance/heat yields and actual performance levels pay off correctly. And they HAVE TO BEAT THE OVERHEATING ISSUES or else come clean on the overheat slow down curve that their version of 10nm actually has to have so the chipsets can be correctly redesigned.
Intel finally had to take this path with their overheating 14nm chipsets, finally having to develop correct specification curves for cooling and performance that actually reflected what they were really actually shipping.
Intel NEVER COULD FIX their 14nm overheating slow down issues, they simply spec'd them correctly in the published product specs. And they correctly spec'd some serious chip cooling as part of the performance curve sales package.
Intel now markets 14nm as a battery saver, not as a performance improvement.10nm ARM may have to do the same. A possible dodge is to go massively multi-core at the ARM 10nm design level,
to keep each individual core's temperature down at a reasonable heat level.
A part of this trick will be the supporting ARM bios software, able to rapidly switch out the tasks assigned to an overheated core to a cooler core. Since many phone tasks are only 1-2 threads deep, a 6-8 core chipset could move things around to cooler cores in a round robin fashion to keep the throughput up higher.
Good chipset cooling DESIGN REQUIREMENTS would be needed as well.
========================================
http://www.ibtimes.co.uk/xiaomi-flagship-smartphones-pack-mediatek-chip-10-co...Late breaking design news from Mediatek and Qualcomm as they try to design some massively multicore chipsets that will work well on into 2016 and 2017 and be able to bump down the lithography shrinks for those years ......
"Qualcomm, in contrary, is also said to be planning for a new SoC featuring the deca-core processor dubbed Snapdragon 818, in a bid to compete with MediaTek's Helio X20. Qualcomm's upcoming chip is expected to support LTE-Cat 10 speeds.
Designed to be used in top-notch mobile devices, the Helio X20 is the world's first mobile processor incorporating Tri-Cluster CPU architecture, with 10 processing cores. The company has confirmed that the chipset would be ready by third quarter of this year and available in consumer products by the end of the year."Qualcomm has a 4 core chipset that is a horsepower chipset, but it sounds like it won't translate well at 10nm with the thermal constraints, so Qualcomm is going to go massively multicore to beat the heat .....