Battery-life claims never seems to line up with reality. You’d think testing battery life would be straightforward, but benchmark results rarely jibe with real-world results—in part, because there are an infinite number of potential workloads (each tapping power differently), and battery life decays over time. Both Intel and AMD make mobile CPU platforms designed for low power consumption, but due to the massive number of variables involved, I’ve found it nearly impossible to determine which architecture sucks the least juice.
Think about it. There’s a lot of hardware in a laptop that can affect battery life besides the CPU and the battery itself: the LCD screen and backlight, the optical and hard drives, the GPU, chipset, and memory config—to name just a few. The upshot is that if you want to fairly compare Intel and AMD hardware, you really need to test what we’ll call core power draw, isolating all the other variables. There are just a handful of ways to do this fairly, and each comes with its own problems.
The best approach would be to test notebooks that are identical save their chipset and CPU. Indeed, if there were a notebook family with the same battery, display, drives, etcetera, that came in Intel and AMD flavors, that would be perfect. We would be able to test performance and battery life (and the relationship between the two) for results relevant to both test machines as well as the broader AMD-versus-Intel debate. This would be the best-case scenario, but unfortunately, these machines don’t exist.
A slightly more realistic option would be to get reference boards from AMD and Intel and put them in a machine with dual power supplies. The mobo, CPU, GPU, and memory would draw from one supply, and everything else would pull from the other. With this config in place, we’d simply measure the power draw of the PSU connected to the motherboard and the benchmark results of both systems.
This is a totally artificial test, but it perfectly isolates the stuff we want to test. That said, were we to go down this path, we could give readers the wrong impression that the CPU, chipset, and GPU are the only things that matter for battery life. If a reader looks at the benchmarks and thinks, “Oh, vendor A has better battery life,” he’s going to be pissed when he buys a big desktop replacement with the winning architecture, but also a 17-inch panel and a puny 3-cell battery, and gets just 30 minutes off a full charge.
So, you see the dilemma. This is why I always say that battery-life testing is a trap. Either I test in a way that actually measures the performance of the CPU, chipset, and GPU, or I give readers actionable buying advice. But I don’t think I can do both at the same time. What do you guys think? Write an email, or let me know via Twitter ( http://twitter.com/willsmith ).