That has more to do with browser SW becomming more efficient, GPU rendering and threading, in particular. Otherwise, you are looking at some rather comparible architectures. If you WERE to compare them.... in terms of compute, the iPad 2 SoC gets around ~15 GFLOPS in theory? The lowest end IGP currently in a modern computer, the HD6250, gets about 44GFLOPS. For another, unrealistic reference, the HD6990m (the mobile version of the HD6870 desktop card), gets 1.6 TFLOPS, or 1,600 GFLOPS. The Fastest desktop single GPU? the HD7970 gets a wonton 3.7 TFLOPS.
However, that is not directly comparible, as the HD6250 has more memory bandwidth (though the SGX is TBR to help negate that sizeable advantage...), MUCH improved texture filtering systems, and much better geometry performance then the raw compute power difference would suggest.
In terms of percieved performance? Probably close. In terms of actual performance? That was the lowest end IGP, and it soundly trounces the A5 GPU in all but power consumption (even so, the HD6250 has a more power efficient vesion that is on a faster chip, too!)