There will be some minor, probably not even noticeable, performance hit at startup. After that don't even worry about it. .NET in all its incarnations has been optimized for years.
Most advice talks about splitting assemblies based on functionality. This has been my personal experience. If you're splitting (or not) for performance reasons rather than architectural reasons, you most likely have a bigger issue.
If there is no specific driver to split your assembly such as reuse or performance, don't do it. Laziness is useful in a project. If you start splitting with no real reason you increase accidental complexity by, for example, introducing potential versioning issues and having extra artifacts to deploy/maintain.