Only slightly related, but GPUs also have such instructions (exposed as InterlockedMax in HLSL and atomicMax in GLSL and CUDA).