A new rumor from the Korean site DDaily suggests that Samsung will not only make an ARMv8-based 64 bit Exynos 6 chip that will be based on a new process node, but will actually skip a whole process node altogether (20nm), and jump straight to 14nm. According to this rumor, instead of seeing 14/16nm mobile chips for the first time in 2015, as initially planned by foundries like TSMC and Global Foundries, we might see 14nm chips in 2014 from Samsung.
That sounds great. There is however one problem with that. It seems very unlikely to happen next year. The “Exynos 6” part sounds right, since that’s what’s coming after Exynos 5. The 64-bit part also sounds just about right since it’s supposed to arrive next year for all chip makers, but it really depends on timing. If Galaxy S5 is really going to be announced in January, and if it’s arriving soon after that, then a 64 bit chip in it sounds pretty unlikely.
The rumor also says it will feature a big.Little set-up, just like before, but this time with Cortex A53 for the low-end, and Cortex A57 for the high-end, both 64-bit. But again, these weren’t really supposed to arrive until mid-2014 at the earliest, if not fall of 2014. Remember we saw the first Cortex A15 in the fall of 2012 for the first time (Nexus 10), and Cortex A57 is its successor, which means it will arrive about 2 years after that (fall 2014).
But when it comes down to it, the most unbelievable part about this chip is really the “14nm part”. The rumor says Samsung wants to “get back in the foundry business”. I’m not sure exactly what that means since they were doing pretty well about a year ago when even Qualcomm started making its chips in Samsung’s foundries.
TSMC has been gaining a lot of customers lately, including Apple, and Apple will continue to shift its chip making away from Samsung to TSMC. This is most likely what Samsung is referring, too, but that doesn’t really explain the almost miraculous transition from 28nm to 14nm in the same time it takes TSMC and Global Foundries to switch to 20nm.
It makes very little sense, and it could be because whoever is sending out this rumor may be confusing two rumors, one for 2014, and one for 2015. I can definitely see Samsung and others making 14nm or 16nm chips in 2015, but 14nm in 2014 seems too early. I’ll be glad to be proven wrong, though, because that would mean Samsung would be even ahead of Intel to 14nm, since Intel is now rumored to have delayed its 14nm Broadwell chips until late 2014 or early 2015.