Games don't really benefit from SIMD. The ratio of instructions to data in games is usually 1 on 1 so there are very little opportunities for SIMD. Although some games do try to utilize it, the result isn't so hot. So yes, some games do use SSE and SSE2, however, very few of them actually benefit all that much from the usage of it.
When SSE/SSE2 really becomes advantageous is in cases in which the instruction to data ratio is very low, i.e. 1:2 or 1:4. Such is the case with video or 3d rendering, audio work, etc. Such applications usually involve a relatively small set of algorithms being repeated over and over again on a huge amount of data. This is where SIMD can be very useful. SSE (present on the AthlonXP and P4) is very helpful towards such applications as DiVX encoding, etc. SSE2 (present only on the P4) allows operations on 64-bit double-precision FP data, which is used quite often in high end 3D CAD programs. The benefit Lightwave 7.5 and higher receives from its SSE2 optimizations should attest to that. In many rendering scenes, a 1.8a P4 can even surpass a 2.25 Athlon XP (with the 333 FSB) in rendering speed.
"We are Microsoft, resistance is futile." - Bill Gates, 2015.