HERE is more on long-term radioactive dating, but essentially the accuracy is a fraction of the half-life of the isotope you are using, not a specific number of years. So if one dating technique gives a 10% error over 3 billion years, that's a possible error of 300 million years. Obviously, such a technique wouldn't work too well if you want to measure the age of something that is only 30 million years old.
Error margins of radiometric dating are order of magnitude of 1%, which is rather disturbing when trying to find out what is cause and what is effect.
Another important tool for dating of very old stratifocations is paleo-magnetism. As the Earth magnetic field has reversed many times, this is recorded in the rocks and an excellent record of these reversals is available. So many scientists earn their living with investigating sediments for magnetic polarisation, trying to identify those reversals.
Yet another dating method is identificing fossils. Many extinct species like planktonic type of micro-organisms that fossilize easily, did have a short livespan. The Sepkoski Compendium (Bulletin American Paleontology no. 363, Ithaca 2002) contains 36,380 of those "marine genera" (single: genus, a group of closely related species) dated back to 540 million years ago.
For the short term there are loads of additional methods, dendrochronology, varve counting (yearly lake sediment layers), Ice core layers and contents, volcanic glass layers (tephra) of known eruptions, thermo - opto illuminicense, wiggle matching and what have you.