A Beef With Battery Benchmarks

Earlier this year, I remember AMD reps stumping about how laptop battery life gets measured. It's misleading, to say the least. MobileMark07, they said, is a synthetic test that measures idle time before a laptop battery dies. Many vendors use MobileMark07 results because it's a repeatable test, and those are the numbers you see pushed in promotional materials. But is that how you use your laptop? Hell no! You're running apps, watching video, listening to music, firing up the Wi-Fi--you get the idea.

Someone I know recently visited a couple brick-and-mortar stores to try a little social experiment. Playing the dummy, she said she needed a laptop with a long battery life. On every occasion, she got directed to an Intel machine. When I mentioned this to AMD's Leslie Sobon, VP of Worldwide Product Marketing, she: "The way people report numbers needs to change."

Well, obviously, these folks are motivated to find other metrics, since people aren't getting directed to AMD rigs. AMD's Brian Henry told me yesterday that he currently likes to run 3DMark06 (which runs the CPU at about 47 percent) while turning on the Wi-Fi to better reflect what you'd probably do with your machine over the course of a day. Other benchmarking suites try all sorts of things--installing apps, turning on Wi-Fi, downloading files. And yes, PC World has its own suite that we reference in our reviews, PC WorldBench 6--a collection of off-the-shelf apps that hammer a computer into submission. (If you want to learn more about our testing procedures, read "WorldBench 6 Released.") But after hearing about all these different solutions, you can't help but sit and wonder if there is a good single test, or if you need to cook up some massive combination of programs.

DIY Personalized Benchmarks

As I sat in a room filled with fellow laptop reviewers and newshounds, I was feeling a little odd as AMD spokespeople prodded to find out more about our testing methods. But what I found interesting is that the company is thinking about releasing an open-source script that lets people customize their own real-world benchmarks. On one hand, hundreds of different tests flooding the market could be a nightmare. On the other, the idea of being able to create tests to suit my own quirky needs ("Can a machine run 20 different Internet meme videos in Windows at the same time?") sounds pretty sweet.

There is one thing that AMD spokespeople brought up in the past and that Sobon hammered home again yesterday: "Let's at least get to where the cell phone industry is and go by usage model times," she said. If you look at cell phone specs, you'll see numbers for idle and talk time. Apple even gives iPhone battery life stats for playing video and music. Wouldn't that be handy for gauging how long a laptop will really last?

How many times can I watch Titanic before the battery dies--or I stab myself in the eye? How long can you play World of Warcraft at the local coffee shop? How long will your machine run while you're transcoding video? Or playing MP3 mixes?

But even more than real-world benchmarks, I want honest-to-God "all-day computing" without the aid of ridiculous battery slices, oversized 200-cell batteries, and the like. On that front, there's some promise. I'm currently playing with HP's Pavilion DM3 (the Intel version--sorry, AMD) and our PC WorldBench tests are showing it lasts about 8 hours. Not bad--but unfortunately, I don't work an 8-hour day these days. More like 10.

What do you think? Are you happy with the benchmarks currently out there? Are you kind of turned on (in a geeky way) by the notion of an open-source plan like what AMD is considering? More to the point----what do you think needs to be taken into account with benchmark tests that currently isn't? Hit the Comment box below or send e-mail to PC World with "ATTN: Darren (benchmark, shmenchmark--does it work?)" in the subject line.

Need even more nerdity? Follow PC World Senior Writer Darren Gladstone on Twitter (gizmogladstone) for more ramblings on all things tech (and games).

Subscribe to the Power Tips Newsletter

Comments