Exynos5-dual has 2/3 the consumption of the Exynos4-dual, not dual vs quad. The Exynos5-quad (2*A15 + 2*A7) has 1/2 the consumption of Exynos4-quad. And i don't understand how you estimate Exynos4-quad at 5-8Wp, 5Wp has a quad-A9@3.1ghz at 40nm-HP from known company.
http://www.anandtech.com/show/6422/s...s-cortex-a15/7
Samsung Chromebook Power Consumption Comparison
Display Off Idle Kraken (Avg) Kraken (Peak)
Samsung Chromebook 500 (Atom N570) 8.82W 11.4W 12.4W
Samsung Chromebook 303 (Exynos 5 Dual) 4.07W 8.32W 9.27W
On chromebook XE303,Kraken-Idle=power of (Cpu+memory)=4.2w, I estimate 8X256M DDR3L consumes 2w during kraken test, so cpu consumption is around 2.4w@kraken test. And kraken test couldn't utilities 2 core of exynos5250 completely, so let's double the power of exynos5250, that is 5w.
BTW, under load memory consumption: DDR3(1.5v)>DDR3L(1.35v)=DDR3L-RS>DDR4>LPDDR2/3.
When idle, DDR3L-RS < DDR3L.
No, no, no. Don't take all those for real. They measure the entire board that is different from smartphone boards. My personal tests only with a cpu say this: A9-single (13m transistor) @2ghz 65nm-HP =950mWp. The same at 45nm-HP =500mWp, and 40nm-LP =350mWp. The same at 28nm-LPE =200mWp, and for dual core 400mWp. If we assume that an A15 consumes 2x as it is, and 3x with the 32nm-Samsung, then 2.5Wp for the entire SOC with graphics and without memory its fine. For example the dual A15 inside the new iPhone what do you thing it consumes? If it was 5Wp without memory then the device will power off in 15 minutes full use.
It is the microarchitecture that is the issue, not the ISA. It is not that they don't have the knowledge to build it, but that it is not targeted. They can easily add more functional units to the pipeline, create a more aggressive cache, and add more hardware techniques that increases performance, but all that comes at the cost of more power which no current company is interested it.