OK, I did some test bursts on the Nikon P510;
The first thing I discovered is you cannot rely on the EXIF data!
If I shoot a burst of images, the buffer fills with 5 Large, Fine jpgs.
At a shutter speed of 1/2000s, this is all over in about 1 second, but the EXIF data assigns them consecutive timestamps one second apart. This means I can only be guided by the first timestamp in a sequence to determine buffer clearance (i.e. card write) speed - and then not numerically (insufficient resolution), only for comparison.
The results, using the four cards available today, showing the number of seconds between the first image in each of two sets of five, shot as close together as possible*, are;
32GB C10 card; 14.5 seconds
8GB C2 card; 7 seconds
4GB C2 card; 7.5 seconds
4GB C6 card; 6 seconds
the .5 second figures arose as I averaged two pairs (e.g. 14 on one and 15 on another = 14.5)
* You cannot shoot one burst and shoot second without, sometime in the buffer emptying process, lifting off the shutter button and re-pressing, this was the most consistent way to test.
Now I am intrigued as to whether the Canon S100, which although I don't use it for bursts, is also affected by the "larger memory space, longer write time", issue. So I may test that in a similar manner, mainly to discover whether the 32GB C10 I now have 'spare' is better off in that camera than the 4GB C6 I was using.
As they say; watch this space ...