16kb is quite big for an embedded processor actually, as most don't have any at all
I think you are comparing a low-power embedded processor with a desktop processor. Sure, Pentium's and AMD chips can have 512KB, 1MB & 2MB caches, but look at the power consumption: An ARM chip likes the one's in the GP32 & GP2X produce less than a watt of waste power, whilst the AMD and Intel processors produces 35 watts for the smallest, mobile, low power version.