Performance difference between Multiple Objects per Mesh and Large Vertex (Polygon) Counts., is subjective.
More Objects means more State Changes Per Frame (1 Per Object in Mesh)… but this is primarily CPU Performance Dependant, and as DBP doesn't use Multi-Core Processing; this will likely be offset by the Graphics Driver offloading to Unused Cores; where-as Higher Polygon (Vertex) Counts rely on Higher GPU ROP Counts., so will affect Framerate more directly (esp. AMD Hardware that tends to have Few Geometry Pipelines) but then this is only the case if we're talking about fairly extreme Polycounts on Modern GPUs.
With this said in both cases the cost is extremely low for either... so unless we're talking Hundreds of Thousands of instances (which you'll more likely run into a CPU Bottleneck with DBP For..Next Loops) chances are neither will have any real major effect on Performance.
You're still however not explaining WHAT exactly you're trying / want to do.
There could be a completely different approach that might be more useful.
Especially as strictly speaking while sure, each Object might have a State Change (assuming different inputs) if you're using an "Uber" Shader as it were., that's not particularly complex per se, but is quite large because it has multiple techniques; or you want to simply have different Colour Variants (which could be done via Vertex Colours blended with the Texture Sample) then it would still be the same State and thus merely cost what an Instant would instead of another Draw Dispatch.
So, yeah; if you could explain what you want to do better it'd help.