Note: Bing varied widely, much more so than Google. First result was 384 ms. second was 320 ms. third was 150 ms. most of the rest were 130-160 ms, but I had another >300 ms. GMaps tended to stay right around 150-160 ms.

GMaps: Av 150.6 ms = 106.2 kb/s
Bing: Av 142.4 ms = 112.4 kb/s

5 refreshes of each:

Google: 147, 144, 137, 147, 178
Bing: 97, 82, 215, 219, 99

NOTE: The browser cache must be cleared between tile fetches because MS is setting Etag and Last-Modified headers while Google does not. Consequently, my initial tests showed the transfer times for subsequent loads from the MS servers to be artificially low because the map tile was being loaded from the browser’s cache (after receiving a 304 response from the server).

If Firebug reports “1 request 16 KB (16 KB from cache) 42 ms” instead of “1 request 16 KB 488 ms”, you should not factor that time into your average speed.

As for needing to compare tiles at the same location and zoom level, as we are just talking about download speed the only thing that might effect results is the file size of each tile and we have made sure both are 16kb.

The time of day and ISP shouldn’t matter also as long as both tests (google and bing) are conducted at almost the same time

Should you not be comparing tiles which show the same place and zoom level? Should you not have suggested an average of five or ten downloads? The Bing result varied wildly (slowest was 161ms). It’s conceivable that download speed will vary with time of day and ISP as well.