I was always curious to what effect different APNs would have on speeds on AT&T, specifically, to see if any of them were given data priority over the other. I tested wap.cingular, phone, and pta APNs and plotted them to see if there was any speed differences between. While there are actual speed differences between wap.cingular and phone/pta due wap.cingular being limited to HSDPA 7.2, I don't think I'll run into them as testing in congested NYC puts me significantly below that threshold. This isn't the most rigorous experimental method to test this, but given the duration it takes to complete all these speed tests (41 in total), this should be enough to satisfy my curiosity.

I made a few other charts to investigate if there was anything particularly interesting in the data. Some of the interesting things that I found were that download and upload speeds weren't correlated at all, that speeds and latency were also not correlated, and that there was only two levels of latency. In the plot comparing APNs, you can see on average the "phone" APN does the best, but its standard deviation is so large that the difference isn't really significant. Subjectively though, phone and pta were definitely faster than wap.cingular, which also timed out more frequently than the other two.

If you want to see the actual raw data, you can see it at: https://docs.google.com/spreadsheet/ccc?key=0AitZVXxNhiYXdEZaZmhhTWNPdkM5R3YzN3h3N2hHUmc

The charts aren't the greatest, but that's the limitation of Google Docs. I was hoping that I could generate the APN comparison using the standard deviation as error bars instead of creating an extra bar for it.

Feel free to copy it and make more plots from it or use it as inspiration to do your own more rigorous or region-specific test.

tl;dr: APNs don't really seem to make a difference if you're under 7.2 Mbps.
4 Photos - View album
Shared publiclyView activity