<span style="font-size: medium">I believe that any of these single shot, max speed memory devices is a complete waste of time.
It will depend a bit on the frame rate. They work by calculating the position, then calculating another position, then calculating the distance between those 2 positions and dividing by the time interval to get the average speed.

If the time interval is very short, then the two positions will be close together and the line between them will be short and almost straight. The trouble is that the positions have a random error. Is the positional error plus or minus 1 metre or 5 metres or what? So if position A is calculated 1 metre too far forward, and position B 1 metre too far back, then the distance error is 2 metres and if the time interval is very short you get a large speed error. Too fast, or too slow.

If the time interval is very long then the position error doesn't matter so much, but in a turn or loop the distance between the two points (straight line) will beshorter than the flight path, so the speed calculated will be wrong (too slow).

The only valid speed data you can obtain from GPS would be a continuous graph of speed (ground speed obviously), so that you can spooth out the spikes and dips due to position errors.

I have used an Eagletree GPS add-on to their data recorder and obtained some useful data. The data can be downloaded as a continuous stream that can be graphed. You can also download heading data (so you know which bits are upwind and which downwind) and if memory serves you can even download the number of satelites locked.
I found that the graph of groundspeed had many spikes and dips, and if I turned with too much bank the number of satelites used dropped, while if I rolled or looped I lost the signal altogether. Even a tight turn could lose the satelites.

You can only believe GPS speed data if it is part of a continuous data stream.
A single maximum spike is laughable nonsense.</span>