I thought it was illegal (in the UK so therefore Europe too?) for a speedo to over read? I've never previously had a vehicle that didn't say I was going faster than the gps did.
Considering how much better the gps receivers currently are compared to only a few years ago with regard to speed of hook up even in less than ideal conditions, and the fact that the speed is just a simple calculation (it would be like checking a calculator to see if when you typed 2 * 2 it really was 4, how would it occasionally produce 4.167253 as the answer?), assuming you are receiving a signal for it to work from and aren't at the north pole the GPS would have to be right? It stops giving a speed when you go through tunnels etc, so it doesn't guess, but it does interpolate between fixes giving a continuously updated average based on a lot of readings and calcs. Driving through a forest may give a few hiccoughs but they'd be obvious prior to either back to normal or a no signal scenario. I can't imagine how gps could be contiuously a wee bit out for driving purposes.
Weirdly, although I have oversize tyres and a speedo not corrected to suit, the zumo matches the odometer better than one percent, but the speedo is over at slow speeds and under at high speeds (speedo and odometer driven from the same drive).
Slightly O/T, a while back there was an earhtworks construction contract in Cornwall. GPS was being used to control the machinery to create whatever it was to be built. At the time you would position a highly accurate surveying receiver in a safe stable position reasonably close to where you were working and leave it taking measurements contiuously, while you used a connected "slave" unit that also received GPS signals, but used the info from the main unit to acheive a better fix. You could then have this slave unit tracking a bulldozer for instance, which would have inside it surveying software with a CAD model of the job. The driver would decide where to drive, but the depth that the blade would add or remove was determined by the software, which I suppose you could tweak to improve fuel efficiency to an optimum depth per pass for that machine on that soil etc etc.
Anyhow, There appeared to be errors appearing, presumably someone had checked what the GPS was doing compared to a known height datum/benchmark, and found height differences of 100mm plus. This could mean a big cost to someone, that depth on a few miles of roadway is a big volume to pay for! I'm sure the person operating the GPS control network was getting quite concerned that he'd made a cock up somewhere, but couldn't find anything wrong in what he'd done.
After a bit of investigating, they discovered that the GPS readings were going slowly up and down throughout every day. Weird.
It turned out that this movement was mirroring the tide coming in and out. Cornwall being a long thin promentary and surrounded by the shallow sea on the continental shelf, when the tide came in the extra water on top of the shelf pushed it and therefore Cornwall down by more than 100mm twice a day! Until you had something remote and seperate in the shape of a satellite or two to measure from no-one was aware this was happening!
Sorry if that was just tooooo boring!!