AMD chief marketing officer and senior VP Nigel Dessau admits that laptop and mobile phone makers give out relatively meaningless numbers when trying to convey to users what kind of battery life they can expect. He suggests that more reasonable tests be developed and used to make the information more reliable.His discussion largely centers on PCs, which Advanced Micro Devices has a vested interest in, but much of the logic would equally apply to smartphones, which are after all, just small computers.

Most PC battery time metrics are achieved by looking at how long the battery lasted running a benchmark called MobileMark® 2007 (MMO7). This is a rating of battery life when your PC is running on average less than 5% utilized -- or fundamentally idle. Most PC makers don't even turn Wi-Fi on for this test. Is this realistic based on how you use your PC?

This explains why Walt Mossberg has his own battery test, where "I turn off all power-saving features, crank up the screen brightness, turn on the Wi-Fi, and play a continuous loop of music."

I used to think battery-life numbers for devices were sort of like EPA fuel mileage estimates for your car: No one actually gets those numbers after driving the car off the lot, but they are consistently applied. If car A gets 24 mpg on the highway and car B says it gets 35 mpg on the highway, you can be reasonably assured that car B will get 25% to 30% better mileage. You also can be sure that the EPA is assuming you are driving the car, not rolling it down a hill the length of the state of Tennessee with the engine idling, which would seem to me to be the equivalent of the MM07 test mentioned above.

Any PC that has Wi-Fi should be tested with that feature on. That's how the average user would use it. For smartphones, someone should be using it, where the screen is on and the cellular radio is working. Because there are generally two types of smartphone users, I'd like to see two battery ratings. The first is pure voice usage: Make a bunch of 30-minute calls, letting the screen shut off as normal until the battery dies. The second test is data: Fire up the device's browser and browse the Internet until the phone dies.

I know both of those tests would likely give lower battery life numbers than manufacturers give out now, but that's OK. The tests that AMD is proposing show numbers 42% to 64% lower than current tests show, but at least the tests are closer to simulating real-life usage. It gives prospective customers better information and, once purchased, the customers won't feel the device isn't performing up to specs -- unless they happen to use it in such a way that utilizes less than 5% of the machine.

There are still an incredible number of variables. Smartphones have Wi-Fi, Bluetooth, touch and nontouch screens, external memory cards and internal flash memory, varying screen sizes and brightness levels, and much more. Still, a full-on usage scenario would give the user a better idea of how much "on time" the device would have. Right now, we are typically given standby time and talk time. What smartphone is ever on standby? These devices check e-mail in the background constantly. Depending on what you have installed, you might be periodically updating RSS feeds and checking multiple e-mail accounts. More and more people have apps like Twitter, Google Latitude, and instant messaging clients that are on and running all of the time, even if the screen is off and the phone isn't being actively used.

Let's hope Mr. Dessau makes some progress getting other manufacturers to use more reliable battery tests for mobile computers and that the cell phone industry will take note and give us better tests for modern smartphones that more closely mimic their real-world usage patterns.