Snow on sea ice is an essential climate variable due to its high albedo and insulating capability. In addition, snow thickness estimates are needed to calculate sea ice freeboard and thickness (SIT) from radar altimetry measurements and uncertainties in snow depth contribute significantly to the resulting SIT uncertainty. In recent years, progress has been made observing or estimating winter snow depth. However, very few snow depth products exist year-round, i.e. including the summer melt season, despite considerable snow remaining on the ice in early summer. This study is the first to assess snow depth estimates year-round with a focus on the so far unexplored melt season. We intercompare estimates (a) based on in situ data from the Warren climatology and (b) two modified Warren climatologies using either ice age data or ice charts, (c) modeled snow depth from SnowModel-LG using two different forcings, (d) a climatology of SnowModel-LG, (e) a combination of the modified Warren climatology and the SnowModel-LG climatology, and (f) finally we also, for the first time, explore the use of dual-frequency altimetry (KuLa) to derive snow depth estimates year-round. We compare the snow depth products against buoy observations of snow depth and airborne measurements of SIT, as well as mooring measurements of sea ice draft together with CryoSat-2 freeboard data. We find that the MERRA-2 forced version of SnowModel-LG agrees better with validation data during the melt season, because the ERA-5 forced version melts the snow much faster than any other product. When a Warren climatology is used, we recommend using the modified version and to strictly only use it in the Central Arctic, where it is constrained by measurements. For KuLa we find that realistic results extend into summer. However, some regions retain snow throughout summer, which likely indicates residual negative biases in CryoSat-2 freeboards and/or positive biases in the ICESat-2 freeboard data.