This is something I did a few years back; it was my third-year project for my degree. A bit different to most of what's seen round here, so hopefully it will be of interest.
For the topography, I used vector data and a DEM from the Ordnance Survey. This was available to me through my university's subscription; private individuals can get similar data but for a charge.
The geological observations were made over a six week period, and were mostly observations of rock type and structural data at points located by GPS. These were also plotted in the field on a number of A4 card sheets of topographic maps.
("Structural data" is the angle that the rocks are at. Sedimentary rocks are initially deposited flat, then mountain-building can tilt them. Wikpedia has an explanation: http://en.wikipedia.org/wiki/Strike_and_dip )
The map itself was created in ArcGIS. I entered all the points into a spreadsheet, with their rock types, data, and also the error recorded on the GPS. Then in Arc, I loaded scans of my field slips, and the topographic data. On that I plotted the points from the spreadsheet as boxes, their size corresponding to the GPS error, with a small circle in the centre coloured according to the rock type. With that as my visual aid, I was able to draw in the contacts between the different rock types. The field slips and the boxes representing my data are of course not seen in the finished map.
The structural data was plotted using Arc's ability to rotate a symbol according to attribute data, and automatically label the symbol.
Once the map was completed, it was saved as a raster image which was then put in an Inkscape document. (I was unable to get either Arc or Inkscape to produce vector graphics in a format the other understood.)
For the cross sections, I used Arc to create topographic profiles, which I then took into Inkscape, and drew the cross sections. The generalised vertical section, which also acts as the key to the rock types, was also done in Inkscape. Finally, the whole thing was composited together, again in Inkscape.
In terms of stylistic aspects, I sought to follow what "real" geological maps produced by the British Geological Survey do. Topographic information is present but in the "background"; the emphasis is on the geology. Colours are moderately saturated, rather than vibrant or pastels. The symbols used for dips and strikes, the use of thicker lines for faults than for contacts, and the use of different styles of dashing depending on how certain the data is, are all pretty much as standard practice.
The cross-sections are lined up with the map and at the same scale to help visualise what's going on. (The East-West section lines up exactly, but not the North-South one since that's at an angle.)
Projection is the Ordnance Survey National Grid, what the topographic souce data was in and the de-facto standard for maps in the United Kingdom. The scale of 1:10,000 is fairly common for use in field mapping and is thus used on the final map too. The intended print size is somewhat larger than A2, but smaller than A1; it's the size needed for all the stuff, and not a standard paper size.
The creation process was by no means easy, but I had relatively few problems with the software. I was one of a minority of my class to use ArcGIS; most people used Inkscape, at that time not very mature software, and I heard of a lot more problems with that approach, perhaps due to the complexity of the projects (in terms of many layers and features). I was fortunate to avoid the arduous task of tracing contour lines, since I had vector data from the Ordnance Survey for that. While I don't consider my map perfect, the main problems are due to so-so fieldwork in the first place. (Getting rained on a lot was expected since I did pick one of the wettest parts of the UK, but all the rocks looking almost the same was something I didn't expect and made things a right pain.)
A few tweaks have been made for the version posted here. My real name has been removed, a couple of errors fixed (and a couple more not!), and the image resolution reduced to keep the forum's attachment system happy.
ArcGIS is very expensive software, so not really suitable for hobbyists. QGIS, a Free Software GIS program, has advanced a lot since when I was making this map, and is now I believe capable of doing the same things.