The process of momentum and energy transfer from a massive body moving through a background medium, known as dynamical friction (DF), is key to our understanding of many astrophysical systems. We present a series of high-resolution simulations of gaseous DF using Lagrangian hydrodynamics solvers, in the state-of-the-art multi-physics code, GIZMO. The numerical setup is chosen to allow direct comparison to analytic predictions for DF in the range of Mach 0.2 ≤ M ≤ 3. We investigate, in detail, the DF drag force, the radial structure of the wake, and their time evolution. The subsonic forces are shown to be well resolved, except at Mach numbers close to M = 1. The supersonic cases, close to M = 1, significantly undershoot the predicted force. We find that for scenarios with 0.7 < M < 2, between 10% − 50% of the expected DF force is missing. The origin of this deficit is mostly related to the wake structure close to the perturber, where the density profile of the Mach cone face shows significant smoothing, which does not improve with time. The spherical expanding perturbation of the medium is captured well by the hydro-scheme, but it is the sharp density structure, at the transition from Mach cone to average density, that introduces the mismatch. However, we find a general improvement of the force deficit with time, though significant differences remain, in agreement with other studies. This is due to (1) the structure of the far field wake (𝑟 ≥ 4𝑟 𝑠 , with 𝑟 𝑠 as the gravitational smoothing scale) being reproduced well, and (2) the fraction of total drag from the far field wake increasing with time. Dark matter sub-haloes, in typical cosmological simulations, occupy parameters similar to those tested here, suggesting that the DF which these sub-haloes experience is significantly underestimated, and hence their merger rate. Dynamical friction is a relevant benchmark and should be included as one of the standard hydro test for astrophysical simulations, since it is critical to our understanding of many systems.