Electromagnetic theory predicts that the optimal value of the localized plasmon resonance (LPR) wavelength for the maximal SERS enhancement factor (EF) is half the sum of the laser and Raman wavelengths. For small Raman shifts, the theoretical EF scales as the fourth power of the local field. However, experimental data often disagree with these theoretical conclusions, leaving the question of choosing the optimal plasmon resonance for the maximal SERS signal unresolved. Here, we present experimental data for gold nanorods (AuNRs), gold nanotriangles (AuNTs), and gold nanostars (AuNSTs) simulating 1D, 2D, and 3D plasmonic nanostructures, respectively. The LPR wavelengths were tuned by chemical etching within 580-1020 nm at a constant concentration of the particles. The particles were functionalized with Cy7.5 and NBT, and the dependence of the intensity at 940 cm-1 (Cy7.5) and 1343 cm-1 (NBT) on the LPR wavelength was examined for laser wavelengths of 633 nm and 785 nm. The electromagnetic SERS EFs were calculated by averaging the product of the local field intensities at the laser and Raman wavelengths over the particle surface and their random orientations. The calculated SERS plasmonic profiles were redshifted compared to the laser wavelength. For 785 nm excitation, the calculated EFs were five to seven times higher than those for 633 nm excitation. With AuNR@Cy7.5 and AuNT@ Cy7.5, the experimental SERS was 35-fold stronger than it was with NBT-functionalized particles, but with AuNST@Cy7.5 and AuNST@NBT, the SERS responses were similar. With all nanoparticles tested, the SERS plasmonic profiles after 785 nm excitation were slightly blue-shifted, as compared with the laser wavelength, possibly owing to the inner filter effect. After 633 nm excitation, the SERS profiles were red-shifted, in agreement with EM theory. In all cases, the plasmonic EF profiles were much broadened compared to the calculated ones and did not follow the four-power law.