Yes, I think it's just a matter of multiplying the vertex position with a transformation matrix built from the bone(s) position/rotation. Since the GPU is much better at matrix multiplications than the CPU, this usually results in a performance increase (unless your application is GPU bound).