Directed energy deposition builds three dimensional components by melting and solidifying metallic powders. Most numerical studies rely on Reynolds averaged Navier–Stokes (RANS) models[1], which cannot resolve instantaneous turbulent fluctuations and may bias predictions of flow driven powder transport. The substrate is also frequently neglected, even though it can induce near wall recirculation and thereby alter particle deposition. This study validates the importance of high fidelity flow modeling and explicit substrate representation for predicting powder transport, deposition, and cooling. We compare RANS and large eddy simulation (LES) for the gas phase flow using turbulence statistics including turbulent kinetic energy and its production and dissipation rates. Three inflow configurations are considered, including a baseline case with carrier and shielding gases, a shielding gas only case, and a no inflow case without gas injection. Deposition is quantified using the spatial distribution of particle numbers, while particle focusing is measured by the number of particles entering the melt pool region. Convective effects are evaluated by computing the Nusselt number from the wall heat flux. RANS is over dissipative, producing attenuated turbulence levels and altered flow structures relative to LES. With the substrate included, the baseline inflow generates near wall recirculation that reduces particle focusing by approximately 30 percent compared with the no inflow case. Nusselt numbers differ by about 10 percent among the inflow configurations, with the lowest value in the no inflow case and the highest value in the baseline case. Overall, LES together with explicit modeling of the substrate is necessary to capture the dominant deposition mechanisms. By contrast, convective effects on particle temperature and substrate cooling are comparatively weak, suggesting that the energy equation may be omitted when thermal convection is not of primary interest.