This work presents a detailed and novel ash deposition model predicting early deposit formation validated with measurements conducted in a full-scale suspension fired boiler considering inertial impaction, eddy impaction, thermophoresis and condensation as deposition mechanisms. A cooled deposition probe, exposed for one week in the furnace region, is modeled using an isolated tube and Large Eddy Simulations. The flow field is validated against literature data using isothermal and non-isothermal cases. Fly ash is described using measured particle size distribution and CCSEM data. Particles with various chemical compositions are injected in the domain and the role of thermophoresis on impaction and sticking efficiency is investigated. The ash deposition model uses a detailed sticking criterion based on the particle kinetic energy, the impact angle and its viscosity. The model is able to predict an increased iron content in the initial deposit layer. Results indicate a high sensitivity towards the viscosity model with best agreement for the model of Senior and Srinivasachar, who implemented a low temperature regime. This approach can predict selective deposition of certain ash particles with low viscosity values such as Fe-rich particles or particles with low kinetic energy, e.g. small Al-Si particles. Depending on the selected model, thermophoresis increased impaction efficiency for small particles by three to four orders of magnitude, and accounted for 5 up to 75% of deposit formation. Condensation on the tube surface is not observed at given temperatures and vapor concentrations. It is shown that boundary effects and particle cooling are crucial effects during early stages of deposit formation.