How much is too much? Upgrade your notebook without going over the line

How we tested

To find out how system memory and hard-drive choices affect performance and battery life, I used a Fujitsu LifeBook T2010 system. The 3.6-pound tablet has a 1.2-GHz Intel Core 2 Duo processor and uses Microsoft's Windows XP Tablet Edition, which is a variation of XP Professional.

I used Windows XP instead of Windows Vista because XP is still the most common operating system on laptops even though Microsoft is phasing it out. And, from a more practical point of view, not all the system benchmark tests I used support Vista. As I mentioned above, Vista's system requirements are higher than those of XP, so if you buy a Vista laptop, plan accordingly.

I used 512MB, 1GB and 2GB RAM modules to vary the amount of system memory from 512MB to 4GB. I tested with both a 60GB Fujitsu MHW2060 hard drive, which has a single disk that spins at 5,400 rpm, and a Samsung 32GB SSD storage module.

The variable I didn't control was graphics memory. The T2010 uses an Intel GMA X3100 graphics processor that automatically apportions system memory to create the images shown on the display. With higher RAM levels, the system allocates more system memory -- up to 384MB -- to graphics. The good news is that, with the same amount of RAM and software running, the graphics chip consistently uses the same amount of memory, so the comparisons are accurate.

To measure battery life, I fully charged the system's 5,800-milli-amp hour battery and, using my lab's Wi-Fi network, I set Internet Explorer to a Web radio station, which provides a constant and repeatable battery drain. With Battery Monitor software running, I unplugged the system and let it run down with the screen and audio adjusted to three-fourths of full brightness and volume. After the system ran out of power, I restarted it and checked Battery Monitor's Log file to find the start and end time of each run. I rounded each result off to the closest five-minute interval.

To measure the notebook's battery life and speed at the different configurations, I used several programs, each of which is downloadable:

• Futuremark's PCMark 05 exercised the system with simulated activities either individually or several at once. It showed not only an overall performance score, but results for memory and hard drive.
• CPUID's CPU-Z confirmed each configuration prior to testing.
• Simpli Software's HD Tach delivered average and burst throughput scores for a hard drive as well as access time.
• Finally, PassMark's BatteryMon gauged power use and helped with battery testing.

Test result numbers are particular to the specific testing software. However, higher scores translate into better performance.

Brian Nadel is a freelance writer based near New York and is the former editor in chief of Mobile Computing & Communications magazine. A 25-year veteran of technology journalism, his work has appeared in Popular Science, PC Magazine and Fortune.