In this talk I present our recent work  on simulating the emergence of the Glasma in the early stages of heavy ion collisions in full 3+1 dimensions. Usually, in the color glass condensate framework incoming nuclei are assumed to be infinitely thin Lorentz-contracted pancakes which leads to boost-invariant rapidity profiles of the resulting energy density after the collision. We break boost invariance by allowing for a finite width of the incoming nuclei along the beam direction and obtain Gaussian-like rapidity profiles already at tree level. The profiles resemble strong coupling results and agree surprisingly well with experimental data of pion multiplicities as obtained at RHIC.
 AI, D. Müller, arXiv:1703.00017