Video: Iron Man
One thing that perpetually bothers me about battery life results generated from video playback workloads, is that they're rarely put into context. Modern SoCs include fixed-function hardware dedicated to decoding. So, a tablet’s run time while watching a movie is heavily dependent on that logic specifically. When we start measuring power consumption in this scenario, we can’t expect correlation with the previous two pages.
And indeed, the Surface doesn’t fare as badly when we play back a transcoded copy of Iron Man using the Windows 8 UI Video app.
Samsung’s implementation of the APQ8060A looks particularly strong, dipping as low as half of a watt less than the ATIV Tab’s Atom Z2760 at certain points. The lines overlap quite a bit, so it’s hard to tell, but the logged numbers show Dell’s XPS 10 using less power than the Atom much of the time, too.
This prompts the question: why do two platforms based on the same SoC demonstrate different battery power behaviors? Aside from the fact that other disparate hardware choices affect consumption, each vendor is able to tune its tablet uniquely for performance or battery life. When dealing with closed-box consumer electronics, you cannot assume that two devices based on the same components will yield identical numbers.
YouTube: Gangnam Style
As if to underscore why one video playback test won’t be indicative of another, we fully cached Gangnam Style on YouTube and logged power use as it streamed.
This blended processing/video decode benchmark spins up the Tegra 3’s cores, spiking the whole platform up over 5 W and well above the competition. The Krait-based Qualcomm SoC incurs a larger hit as well. Both the Samsung ATIV Tab and Dell XPS 10 average higher consumption than the Atom-based Acer Iconia W510, which is sort of in line with what we saw from the more demanding Tom’s Hardware homepage test.
The implication is that, at idle, the ARM-based designs exhibit superior power use. As load increases, Atom appears to draw even, and eventually operates more efficiently.