Also, it should be noted that the Nano architecture provides for additional power saving options over the C7-M. Still, if you are concerned about battery life, you are in Intel's camp. Their stated goal in the Atom was to improve battery life and keep performance steady. Via's stated goal in the Nano was the opposite - better performance and keep battery drain/life steady. I vote for faster, myself.
The question of how to optimize for battery life is definitely tricky. The PC Perspective review talks about how Centrino 2 takes the "Hurry up and Get Idle" strategy. I think Via may be trying the same sort of thing with Nano. That is, get the work done as quickly as possible then transition to its lowest power idle state to wait until there is more work. This may be a win if it can transition between power states quickly enough and for long enough. (Obviously, this is a continuum. The U-series Nanos, because they're lower max power and lower performing, are closer to Atom in their power strategy.)
Here's another Nano vs. Atom comparison article:
http://techreport.com/articles.x/15204 (Note: This article benchmarks applications that seem unlikely on a mobile device. They admit as much. However, they're testing the desktop versions of the processors.)
I did a quick scan. Not surprisingly, like the other articles, they show that Atom and Nano consume approximately the same amount of energy for a given task, Nano just does it over a shorter time. (Which one consumes slightly less energy depends on which article you're reading.) Keep in mind though that this is the Nano desktop part with the 25W TDP. The U-series parts won't jump 20W in power consumption when heavily loaded. Of course, it's also unlikely to get as much work done per unit time. If we make the rather dangerous assumption that performance scales linearly with clock speed, we get the conclusion that the U-series parts will likely be more energy efficient.
(i.e., if we assume that a task takes the 1.8 GHz L series part 1 second at 20W, and the 1.3GHz U series part 1.38 seconds at 8W, the latter still only consumed ~11 joules rather than 20. Again, this assumes that performance scales linearly. Big assumption.)
One possible conclusion is that the L series part is too far on one end of the continuum. This isn't really surprising since this is their desktop part. Power consumption probably isn't their prime worry here. How ever well, nor not well, the L series part does, the U series may do better by pulling back towards the "slow and steady" end of the continuum. It obviously does not pull as far back as Atom has.
A more audacious leap to conclusion might be that Atom may have gone too far towards the "slow and steady" side of the continuum. i.e., a computationally more powerful device might actually be more power efficient, if it can get to its lowest power state quickly enough and stay there long enough. Again, keep in mind that this is the desktop version of Atom. It explicitly does not have some of the power-saving bells and whistles that Intel puts in their mobile device version. We don't have enough data to justify any conclusions anyway.
Ultimately, I suspect that it'll still work out that the mobile versions of Nano and Atom will consume approximately the same number of joules to accomplish a given task. I doubt either design is particularly wasteful. The Nano may finish the task more quickly though. (i.e., consume those same number of joules over a shorter period of time. Higher wattage.)
If you are constantly pinning CPU utilization then, the Nano strategy will result in shorter battery life. However, that's not the typical scenario, especially for a mobile device. In the typical scenario, your computer spends much of its time waiting for you. In that case, assuming the two processors have roughly the same idle power, assuming they transition to and from their lowest power states about as quickly, they'll probably consume about the same amount of power. You'll just see your results faster with the Nano. (OS background processes may screw this up. YMMV.)
If you really are constantly pinning CPU utilization, there's no happy solution. I mean, yes, in that scenario, Atom will have consume less power per unit time, but it also will have done less work per unit time. After the battery has drained, you will have gotten about the same amount work done as the Nano. It will have just taken longer. The moral of the story is: don't do that on a battery operated device.
Just to make this point again: This is speculation on my part. A lot of this depends on how quickly they can transition into and out of their lowest power states. I, obviously, don't have access to that sort of information. If I had to guess, I'd say that Atom and Nano will, in general, be comparable. The right part to use will really depend on the specific use. (i.e., Nano is 64 bit, potentially higher performing, pin compatible with C7M. Atom ought to fit into thermally constrained situations where you can't evacuate more than a couple watts of heat. It's dual threaded which may be an advantage in easily parallelizable throughput bound workloads.)
Author
