I am getting similar problems doing the ArcHydro tutorial at step "Build Walls." ArcMap 10.2.2, with ArcHydro 10.2.0.96 for 64-bit. It crashes the first time. If I restart ArcMap and try again, it starts, the status bar fills to 100% in just a couple minutes, and then it hangs, it's been running 20 hours now. The memory allocation goes back and forth between about 70 MB and 1 GB, the CPU is full the entire time. This is a six core i7-4770K 3.5 GHz machine with 32 GB of RAM, Windows 8.1 for 64-bit, so I doubt it really takes anywhere near this long if it were working properly. I had the same problem on a slower Windows 7 laptop (crash and then runs forever on second try), and a co-worker had the same problem on her machine. It's not temp files because it doesn't crash, it just runs forever. So I think that something is wrong with either the tutorial dataset or ArcHydro's ability to handle it.
Early on I eliminated crashes, or at least reduced them, by putting all my layers on the C: drive. Before that I had all the data on the network. Moving the data to my PC not only greatly reduced crashes, it sped the processing time up noticeably.
Thank you, sir. I do have all my layers on C:\ drive, but I’ve also been told that there are problems with grids in particular unless the path to them is really short, so I may move them to something short like C:\workspace just to make sure it’s not something simple like that. I’m just not sure what’s going on. The ArcHydro tutorial, which does work, uses elevation models from NHD+, and I may try that. Until now I’ve been using a DEM that we have always used with Weasel for the same operations (basin characteristics), and I was doing that because we have to use ArcHydro to match the output from Weasel until we can get a different workflow approved by our federal partner. Weasel only works with Arc/Info 10.0 though so it’s time to move to ArcHydro.
I've heard that having different projections can cause problems in geoprocessing that otherwise don't make sense. I've always worked with data that was from our own data managers and therefore on our standard projection, so I've not had to deal with the different projections. So, see if the projections are different in your layers.