This is admittedly confusing although it is discussed in the (very long) chapter on solvation models: 11.2 Chemical Solvent Models‣ Chapter 11 Molecules in Complex Environments: Solvent Models, QM/MM and QM/EFP Features, Density Embedding ‣ Q-Chem 5.4 User’s Manual.
The differences between these two approaches comes down to the definition of the solute cavity. SOLVENT_METHOD = ISOSVP implements the SS(V)PE model developed by Dan Chipman, using a cavity that is defined by an isocontour of the solute’s electron density (as done in Chipman’s papers). That’s a nice, clean, one-parameter cavity definition that I like a lot, but with the important limitation that it lacks analytic gradients. You can consult my review (https://doi.org/10.1002/wcms.1519) for a discussion of why those gradients are challenging to formulate. However, it sounds like this is a non-starter for what you want to do.
Methods that do have gradients include SOLVENT_METHOD = PCM, = SM8, and = SMD. The two SMx models are packaged up with nonelectrostatics terms in addition to electrostatics, and in the SMD case the electrostatics actually come from PCM. (There are also nonelectrostatic interactions for the isodensity SS(V)PE; that model is called CMIRS: https://doi.org/10.1021/acs.jctc.6b00644, but it inherits the lack of gradients.)
Therefore, for geometry optimizations let’s talk about PCM. These methods use union-of-atomic-spheres cavity constructions, with atomic radii that are parameters of the model. Gradients are available. There are several different PCMs, the choice of which is stipulated using the “Theory” keyword in the $pcm input section:
(Other options are iefpcm and cpcm, and note that there is no “=” as in your example.) The differences between these approaches are somewhat subtle; see the review cited above or else here: Redirecting). SS(V)PE and IEF-PCM differ only by symmetrization of certain matrices that makes SS(V)PE about 2x faster (probably not noticeable except for very large molecules as the PCM time is minor overhead), but this speedup sacrifices some formal properties when van der Waals radii are used to construct the cavity. IEF-PCM uses asymmetric matrices but preserves those properties and is what I recommend. However, C-PCM is equivalent to IEF-PCM in the high-dielectric limit and is simpler and faster; “high-dielectric limit” operationally means epsilon >~ 30-35, i.e., acetonitrile or anything more polar than that.
Summary: for geometry optimizations, I recommend SOLVENT_METHOD = PCM with Theory set to either IEF-PCM for nonpolar solvents or C-PCM for polar solvents.
For “dimensions of isodensity cavity”, I assume (?) that you mean just printing out the discretization points that define it. There’s probably a print option to do that; I will poke around the manual and see if I can find it, but if you’re in a hurry check out this section of the manual: 11.2.5 Isodensity Implementation of SS(V)PE‣ 11.2 Chemical Solvent Models ‣ Chapter 11 Molecules in Complex Environments: Solvent Models, QM/MM and QM/EFP Features, Density Embedding ‣ Q-Chem 5.4 User’s Manual. This is where isodensity implementation is described.