Tablet Two Ways: Deconstructing Battery Life Measurements
I started this story by telling you generalizations about battery life are potentially misleading, and then proceeded to make observations about battery life. What gives?
Well, we isolated two usage models that typically come up in our tablet reviews: web browsing and video playback. By measuring battery power over time across four devices under multiple scenarios, it’s possible to see how consumption changes depending on the specific workload. Moreover, we have examples of the placing order changing, again, dependent upon on the workload.
The reason we’re exercising caution should consequently be pretty clear. Reviews need to be explicit about what they’re using to generate battery life numbers. And even then, it’s only possible to draw conclusions about power performance in that one specific scenario. Running more or less demanding test cases will change things up, sometimes quite significantly.
Averaging the power consumption of all five tests suggests that the Qualcomm-powered ATIV Tab and Intel-driven W510 are roughly equivalent when it comes to power draw in web browsing and video playback. However, the individual benchmarks make it clear that specific workloads can exact much different demands. This will become an even more important discussion in the months to come as competition tightens and SoCs become more powerful.
The Tablet Space In 2013
It’s going to be an exciting year for smartphones and tablets, based on what we’ve seen during the first two days of CES.
First, we have Nvidia's Tegra 4, armed with its reference Cortex-A15 cores and newly-separated pixel and vertex shaders. We don’t have power or performance figures for it yet, but it’s easy to imagine a very power-friendly SoC at idle, and a power-hungry chip under even light load. In talking to Nvidia about its Shield gaming device, the company says it wouldn’t be unrealistic to achieve 24 hours of video playback from a 38 Wh battery pack. Even if you factored out all other platform components, the SoC would have to sip less than 1.6 W for this to be possible. That’s less than half of Tegra 3. In order to get the 5-10 hours of gameplay, Tegra 4 would need to run between 3.8 and 7.6 W. We're immensely curious to see if the company can do this.
Then we have the lower-power Atom processors composed of 22 nm transistors and low-voltage Ivy Bridge-based parts, which we might expect to apply pressure from both ends of the power band. Intel claims that its Haswell architecture will be exerting influence in the same space as quad-core Cortex-A15-based devices by the end of the year.
Moreover, Qualcomm impressed us with preliminary numbers on its Snapdragon 800 and 600 SoCs. Although the company’s CES announcement was light on specifics about what it altered in the Krait architecture to get additional performance from the CPU cores, we do know that clock rates are up. Additionally, the Krait 300 inside the Snapdragon 600 now include hardware data prefetch, better branch prediction, improved memory management, and a handful of other tweaks for higher performance in specific tasks. To those enhancements, the Snapdragon 800’s Krait 400 CPU incorporates a redesign to optimize for high-K 28 nm manufacturing, lower memory latency, and faster L2 cache.
At the end of the day, though, cranking performance up 40 or 75% is only meaningful if it can be done without negatively affecting power consumption. We’re not yet certain how Tegra 4 and Snapdragon 800 compare to their predecessors in that regard, but we do find it interesting that Intel is the only company eager to illustrate the power of its designs in a lab setting. By picking up the torch and granularly tracking the consumption of individual subsystems, we hope we can shed some light on the strengths and weaknesses of each respective architecture as they surface in the next generation of tablets and smartphones.
Stay tuned to our CES 2013 landing page as we continue covering the year's biggest consumer electronics trade show.