Perf/Watt is the only thing which matters in the mobile world. Having faster processors is great but sacrificing battery time will make the SoC not competitive.
But like I said, perf/W isn't linear so it's not just a fixed number you can compare. This is going to be especially true for Tegra SoCs with companion cores and big.LITTLE configurations using the cores the same way - perf/W is going to be even less linear when it crosses from the little to big CPU.
I disagree that it's the only thing which matters too, especially if you're only looking at the CPU. Peak perf of the CPU, peak perf and perf/W of other components, especially GPU, additional integration options, price, and yes political standing all matter a lot.
That is not really new. Higher clocks needs higher voltage. And voltage increase power by +%increase^2increase instead of +%increase.
It isn't new, but it's something people should think about more. The voltages you need to reach clocks varies a lot with design and process. Intel deserves a lot of credit for being able to hit 2GHz at the voltages they have, the power consumption of Atom at this point seems very good.
Voltage/frequency isn't necessarily linear either. I think my example is pretty illuminating, look at how much power consumption goes up for that measly 200MHz at the end. Apple kept their 45nm A9s clocked at only 1GHz on tablets and 800MHz on phones, and from this we can see why (it's reasonable to assume that the power consumption was basically the same as on Exynos 42xx, die shots reveal the physical implementation was the same as you'd expect). Samsung was obviously willing to allow huge power consumptions at high clock speeds and let the phones get pretty hot, although the 32nm Exynos 44xx brought a huge improvement.
It really does make me feel like the question that should be asked is, "what's the power consumption like with proper voltage tuning at lower clock speeds", instead of just instantly declaring that the power consumption sucks - what if you get half the power consumption at 1.3GHz? Will it still totally suck then?
It also makes me wonder what power consumption will be like on TSMC 28nm. Particularly on the G process instead of the LP, if we can see such a thing (maybe on Tegra 4?), since that should allow high clock speeds to be hit with lower voltages. Samsung's move to 28nm will help too, although not as much as their move to 32nm did for Exynos 4.
Arm it seems has an army of soldiers spreading misinformation , Which makes since . Apple fanbois can be hardcore . Along with other nameless fans of 1 company or another
Fake information about upcoming products circulates all the time. Really unfair to automatically blame the competing companies for this. Not like Intel fanboys can't be extreme too. But a lot of wrong information probably comes from people misinterpreting or miscommunicating information, and attributing more authority to people than they should have.. passed from person to person it ends up worse at each step. I've seen cases where random forum posts from totally unconnected people end up as sources, and the post was just someone's wishes or estimations posed in an over-confident fashion. Part of the misinformation could also just be from people looking for attention and loving the impact they have.
Facebook
Twitter