A number of older microarchitectures have implemented AVX by passing 256 bit vectors through a 128 bit SSE SIMD unit twice. Effectively, AVX is not faster than SSE on such
