Huawei's Kirin 930 flagship CPU uses enhanced Cortex A53e cores pushed to 2GHz clock speed

Huawei's Kirin 930 flagship CPU uses enhanced Cortex A53e cores pushed to 2GHz clock speed
Huawei is making serious progress with its chipmaking business, coming up with more impressive in-house processors for its smartphones each year. The one that's en vogue for 2015 is the Kirin 930, an eight-core monster that's built on a 16nm process. Previously, the available information suggested that the K930 is based on stock ARM Cortex A53 and Cortex A57 cores, with the most recent reports surprisingly revealing an all-Cortex A53 octa-core configuration at the helm. 

Predictably, some questioning ensued, as the A53 cores are low powered and play second fiddle to the hotter A57 cores in the typical octa-core configuration. However, Huawei chimed in with an interesting explanation. Apparently, the Kirin 930 is based on four regular Cortex A53 cores and four enhanced Cortex A53e cores. The latter appears to be a Huawei creation, and are remarkably different to the traditional Cortex A57 and Cortex A53 cores. For starters, stock A53 cores traditionally run at 1.2GHz, which doesn't cut it for flagship performance. So Huawei pushed them as hot as 2Ghz, which makes them nearly as fast as Cortex A57 cores, but more power-efficient.

According to the company, A57 cores, which are about 56% faster than A53 cores, could eat up as much as 256% more electricity, which results in heat creation and reduced battery life. So Huawei made a careful compromise and ended up with slightly slower, but hopefully more power-conscious cores.


source: MyDrivers.com via GizmoChina

FEATURED VIDEO

26 Comments

1. arch_angel

Posts: 1651; Member since: Feb 20, 2015

great idea since they dont have 14mn processes its a good way to get some great energy efficiency

14. vincelongman

Posts: 5724; Member since: Feb 10, 2013

But how did Samsung's 20nm Exynos 5433 manage to use 4 1.9Ghz A57s and still have no overheating issues?

15. arch_angel

Posts: 1651; Member since: Feb 20, 2015

umm Samsung is a boss lol. they put the chip in a big enough phone that could properly distributed the heat and the devices with that chip are plastic not metal.

16. vincelongman

Posts: 5724; Member since: Feb 10, 2013

The size difference between the Note 4 and One M9 won't make a difference since the One M9 has an aluminium unibody Aluminium absorbs and spreads heat far better than plastic Hence why CPU/GPU heat sinks use Aluminium I think Qualcomm is just inexperienced with big.LITTE

17. arch_angel

Posts: 1651; Member since: Feb 20, 2015

qualcomm just messed up cause there definitely not inexperienced? and metal also retains way more heat than plastic so its not that great the note 4 is one of the coolest running high end phones there is. a metal phone is more prone to overheat than a plastic one. and the size definitely make a difference.

2. carlemillward unregistered

I wonder if it actually is 16nm?

6. luis.d

Posts: 354; Member since: Dec 04, 2013

Yep, TSMC at play.

13. vincelongman

Posts: 5724; Member since: Feb 10, 2013

Technically, no TSMC's 16nm process is still 20nm FEOL and 20nm BEOL Also Samsung's 14nm LPE process is 14nm FEOL and 20nm BEOL Only Intel's 14nm process is 14nm FEOL and 14nm BEOL

18. kelvinkoh0308

Posts: 5; Member since: Mar 28, 2015

Can u tell me why it still can call 16nm?feol and beol are stiil 20nm

19. vincelongman

Posts: 5724; Member since: Feb 10, 2013

They are marketing their 20nm FinFET process as 16nm since it performs as good as a 16nm Planar process They have a "16nm FinFET+" process coming next year, which is rumored to be actually 16nm FinFET proces

20. kelvinkoh0308

Posts: 5; Member since: Mar 28, 2015

if compared to samsung 14nm,samsung is better?

22. vincelongman

Posts: 5724; Member since: Feb 10, 2013

Samsung's 14nm LPE is better than TSMC's 16nm Like I said TSMC's 16nm process is still 20nm FEOL and 20nm BEOL Samsung's 14nm LPE process is a mix of 14nm FEOL and 20nm BEOL Intel's 14nm process is the best out of all of them (fully 14nm) Basically FEOL is the components, e.g. the transistors. Smaller transistors means more performance/efficiency and less size BEOL is interconnecting the components. Smaller BEOL means less space (allows more transistors per nm, for more performance)

21. kelvinkoh0308

Posts: 5; Member since: Mar 28, 2015

And tsmc 16nm is same with the perfomance of the 20nm?

23. vincelongman

Posts: 5724; Member since: Feb 10, 2013

TSMC's 16nm process (20nm FinFET) will performs as good as a 16nm Planar process, better than their 20nm Planar process (used in the A8 and 810)

24. kelvinkoh0308

Posts: 5; Member since: Mar 28, 2015

that means tsmc 16nm is only 20nm +finfet?

25. kelvinkoh0308

Posts: 5; Member since: Mar 28, 2015

that means tsmc 16nm is only 20nm +finfet?

26. vincelongman

Posts: 5724; Member since: Feb 10, 2013

Yea, basically that's it

3. iphonexus

Posts: 130; Member since: Jan 04, 2015

yeah ... I hope that Qualcomm also find ideal solution rather limiting cpu performance ....

4. itsdeepak4u2000

Posts: 3718; Member since: Nov 03, 2012

I really liked their thinking and that's why the Honor 6 Plus heats less than the competition, even better than Note 4.

5. medtxa

Posts: 1655; Member since: Jun 02, 2014

Good job huawei this seems promising, I like it when company able make their own component and innovating.

7. Guaire

Posts: 890; Member since: Oct 15, 2014

They wanted make it cheap. Also I don't believe it's manufactured at 16nm. They said that anywhere? It doesn't seems likely to me choosing very expensive 16nm manufacturing process while they are choosing A53 and Mali-T628 for lower cost. More importantly A53's given die size implying it's more likely 28nm. According to Anandtech, Exynos 5433's single core of A53's die area is 0.7mm2 at 20nm. We know 2GHz A53's rough performance from MT6795's Geekbench scores. Also Snapdragon 615 uses two quad core A53 cluster with Snapdragon 615. So it's not a new approach. Interesting part is about A72. We got a bit more info about it. Seems it's die area smaller than A57 and delivers more performance at same clock speed with less power consumption.

8. TylerGrunter

Posts: 1544; Member since: Feb 16, 2012

Those are lame excuses for the real fact: The single threaded performance will be far away from the A57 cores. The A53 cores have more or less the same performance than then old A9. Not much better or worse. That means: worse than A15, and much worse than A57. Yes, they will produce less heat and occupy less space (saving money for Huawei) but please don't go selling then that the phones are "flagship" killers when they have that silicon on them.

9. kaikuheadhunterz

Posts: 1157; Member since: Jul 18, 2013

Well my only gripe is them not creating a new name for the custom processor

10. jove39

Posts: 2147; Member since: Oct 18, 2011

A53 is 2 wide in-order core A57 & A72 are 3 wide out-of-order cores. Out of order execution hardware consumes lot of power due predictive nature of implementation. Also task scheduler plays major role in efficient implementation. I don't like the idea of dis crediting A57 & A72 cores in comparison to A53. How do we know load scenario was not favorable to in order execution.

11. vuyonc

Posts: 1091; Member since: Feb 24, 2014

Cough cough does the GPU work this time or is it too weak for flagships again cough cough

12. micEvolution

Posts: 4; Member since: Mar 27, 2015

I read something about the last pic that Huawei was able to integrate a base band inside the Kirin 930 by using less 'dies' compared to Apple. Thus, the higher integration density results to more power efficiency and more spaces being saved that the size is equivalent to a 300mah battery.

Latest Stories

This copy is for your personal, non-commercial use only. You can order presentation-ready copies for distribution to your colleagues, clients or customers at https://www.parsintl.com/phonearena or use the Reprints & Permissions tool that appears at the bottom of each web page. Visit https://www.parsintl.com/ for samples and additional information.