Accurate predictions with quantifiable uncertainties are essential to many practical turbulent flow applications exhibiting extreme geometrical complexity and broad ranges of length and time scales. Under-resolved computer simulations are typically unavoidable in such applications, and implicit large-eddy simulation (ILES) often becomes the effective strategy. We focus on ILES initialized with well-characterized 2563 homogeneous isotropic turbulence datasets generated with direct numerical simulation (DNS). ILES is based on the LANL xRAGE code, and solutions are examined as function of resolution for 643, 1283, 2563, and 5123 grids. The ILES performance of new directionally-unsplit high-order numerical hydrodynamics algorithms in xRAGE is examined. Compared to the initial 2563 DNS, we find longer inertial subranges and higher turbulence Re for directional-split 2563 & 5123 xRAGE - attributed to having linked DNS (Navier-Stokes based) solutions to nominally inviscid (higher Re) Euler based ILES solutions. Alternatively for fixed resolution, we find that significantly higher simulated turbulence Re can be achieved with unsplit (vs. split) discretizations.