Use an acceleromter to measure tilt. "Measure" being the operative word you used. A Gyro does not measure tilt, it merely measures rotation rate. If the rate is integrated, then you can tell a relative change in tilt over a given time interval. Keep in mind I worte "relative change in tilt", because the starting state is arbitrary, and with integrated gyro measurements you then have a way to take a previous known tilt, and project forward as time progresses to have an idea of the new tilt. Why all this fuss? because accelerometer measurement quality suffers under the influence of vibration which corrupts the tilt measurement, but the ADXRSxxx gyros are practically immune to vibration. You need to apply proper RC filtering to the accelerometer output to dampen the vibrations, but not induce too much lag. At the same time the gyro is used to track relative changes in the tilt. Using accels and gyros together like this is done because they have complimentary properties. You'll need a fair amopunt of computation power and clean fast code to pull all this off. This is the stuff of a PhD thesis, typically, if you want to do it properly (there are lots and lots of complicating factors to handle).
On the other hand, a MUCH simpler thing to do if centrifugal forces are not at play as in some sensor for your desktop (as opposed to in a moving vehicle), is to simply use a 3-axis accelerometer (like the ADXL-330) and heavily RC filter the outputs to around 9 Hz cutoff. Then it's a matter of relatively simple trig to compute attitudes (tilts). This is what is done in the I-phone as far as I understand that device.

Why do not you sell them, anymore?
These are very good.

So if I want to measure tilt I want a gyro and not an accelerometer, right? I'm pretty sure this is right, but I've seen some projects where people measure tilt with accelerometers. Any help?