I've found the variation from test to test, and site to site, to be incredibly variable. To get any meaningful results is going to require averaging a bunch of tests. And any test is location dependent, so ideally you'd want to repeat them from multiple locations.
Anyway, here are 10 tests each for 3G and 4G, using 2 different test services. I've got full bars on the 3G signal, and only 1 on 4G. Hopefully 4G will get faster as the signal gets stronger. All tests were from the same location, within a span of about 30 minutes.
Code:
mobile.speedtest speedtest.net Android app
Test # [COLOR="Blue"]3G down[/COLOR] [COLOR="Red"]4G down[/COLOR] [COLOR="Blue"]3G down[/COLOR] 3G up [COLOR="Red"]4G down[/COLOR] 4G up
1 928 816 766 737 1408 1041
2 722 1489 1372 674 2400 918
3 688 1381 1312 674 2051 1011
4 543 1283 652 537 1362 1035
5 1012 1282 783 643 1401 797
6 793 1275 739 697 787 855
7 1299 1276 908 637 2111 804
8 1168 1452 1116 697 1740 960
9 1451 1418 937 692 1489 850
10 1396 1404 841 715 2237 874
Sum 10000 13076 9426 6703 16986 9145
Avg [COLOR="Blue"]1000 [/COLOR][COLOR="Red"]1307.6 [/COLOR] [COLOR="Blue"]942.6[/COLOR] 670.3 [COLOR="Red"]1698.6[/COLOR] 914.5
So on the mobile.speedtest site, 4G was, on average 31% faster. On the speedtest.net app, it was 80% faster. How representative is this? No idea.
Oh, I read somewhere that speeds are slower if you have the GPS radio turned on, which I did. I also had BT turned on. Maybe next week I'll try again with only the 3G and 4G radios on.