The single most common complaint I hear is "the GPS took me on a crazy route that I would never have chosen and was not the best/shortest way."

Arguably the most critical components of any turn-by-turn GPS navigation system are the routing engine and map accuracy. The routing engine is the software on a GPS that decides which route the unit should choose to get from point A to point B. Many people falsely assume that simply having the same maps on two GPS devices will result in the same routing choices. Not so, as there is a great deal of math and decision-making involved when calculating the "best" route.

Not surprisingly, this also one of the most difficult and complex aspects for GPS manufacturers to get right, and is the aspect of GPS navigation that is least well understood by consumers.

The two largest mapping data providers are NAVTEQ and Tele Atlas . NAVTEQ is more accurate and complete in North America, so it's preferable to buy a GPS that uses NAVTEQ rather than Tele Atlas (as of this writing). (for more detailed information on both NAVTEQ and Tele Atlas, check out "NAVTEQ vs. Tele Atlas, Which one is Better?")

However, there is a lot more to a routing engine than just the mapping data provider. How a given GPS will determine the "best" route is the result of not only the mapping data, but also how many additional data points, or attributes, are factored into the equation. For example, are certain roads closed after hours or on the weekend? What is the average rate of speed for a given road? Statistically speaking, more accidents happen when making left turns as opposed to right turns -- is the GPS unit factoring that into the route? Each company takes all this data and compiles it in different ways, making unique decisions about what data is more important than others. The number of attributes used varies depending on how much money the manufacturer is willing to spend on their mapping data. For example, a full NAVTEQ map database that includes all available attributes can cost the manufacturer twice as much as a basic set of attributes.

All Magellan and Garmin GPS units use NAVTEQ mapping data, while TomTom uses Tele Atlas. Magellan's Maestro unit ships with 2007 NAVTEQ maps pre-installed, while the Garmin nuvi's NAVTEQ maps are several years old, (although Garmin is planning to release updated maps within the next month or two). TomTom newly updated Tele Atlas maps are also from 2007.

In order to find out which GPS has the best mapping and routing capabilities, I tested the latest and greatest models from Magellan (Maestro 4050), Garmin (nuvi 680), and TomTom (ONE XL). Note that the TomTom ONE XL I tested is the newest version and has TomTom's newly updated maps installed on it..

To accurately test each of the units, I entered three different trips of varying distance in three different regions of the United States.

Our first test took us along the Northeast, starting in Southbury, CT and ending in New York City:

Magellan's Maestro 4050, Garmin's nuvi 680, and TomTom's ONE XL all get us to our destination address, but the Garmin nuvi chooses a slightly faster route, provides more detailed navigation instructions, and is also the only GPS to announce what side of the street our destination address is on.

TomTom's ONE XL faired the worst in this test, routing us in such a way that the destination address is actually on the left side of the street. We would then either have to make an illegal U-turn or go around the block to arrive at 135 Central Park West on our right. TomTom also has less detailed navigation prompts than the Magellan Maestro 4050. TomTom also reports distances in yards instead of feet.

Yahoo! Maps (not shown in the table above) chooses a route identical to Garmin's route, but estimates the trip will take 3 minutes longer (1 hour 22 minutes) than Garmin estimates. Google Maps also chooses the same route as the Garmin nuvi.

So far Garmin's nuvi is in the lead in our routing battle. However, sometimes a GPS that routes perfectly well in one area can perform terribly in another region. Our first test used a trip that was approximately 80 miles in length and took us from Connecticut to New York City.