This page compares the full and reduced spectral resolution versions of RTE-RRTMGP v1.5. The full resolution results are identical to those of v181204 of RTE-RRTMGP.
Longwave
Both versions use 16 longwave bands which almost fit into the 13 "narrow" CKDMIP bands. The full spectral resolution uses 256 g-points, the reduced uses 128 g-points. Evaluation of present-day calculations against the line-by-line reference for the CKDMIP "Evaluation-1" dataset are as follows:
Both are very good models, but we see that the reduced resolution degrades most metrics a little, except for 4-1100 hPa heating rates, which are unchanged. The warming bias in the mid-mesosphere introduced with the reduced spectral resolution is entirely associated with the 630-700 cm-1 band, where CO2 absorption is strongest.
In terms of radiative forcing associated with perturbations to the major greenhouse gases, there is no change for CO2 but slightly less accurate forcings for the other gases:
Full:
Reduced:
Shortwave
Both versions use 14 shortwave bands which almost fit into the 13 "narrow" CKDMIP bands. The full spectral resolution uses 224 g-points, the reduced uses 112 g-points. Evaluation of present-day calculations against the line-by-line reference for the CKDMIP "Evaluation-1" dataset are as follows:
Again the reduced resolution has slightly larger errors, but it is still clearly an excellent model.
In terms of forcing, the reduced-resolution model is a little poorer for very high concentrations of CO2 and CH4. Both models underestimate the N2O forcing, although in the shortwave this gas has a much weaker impact (and was ignored in the earlier RRTMG model).