The answer is that it can go either way. Evolution uses whatever mutations come up, and a mutation for increasing mathematical intelligence could come up in a form that is sex-linked, or in a form that is not. (And we already know that fetal development of the brain is mostly the same for both sexes but does have some little differences in response to sex hormones, so examples of both sex-linked and sex-unlinked genes for something about the brain are already established.) Sometimes the same trait that was sex-linked at first can even become unlinked, or the other way around, if the creation or breaking of the link is beneficial in selection. For example, in Asian elephants, having tusks is a male trait, but in African elephants, both sexes have tusks. So whichever way it was in their last common ancestor, it had to have changed (becoming sex-linked or losing that link) in at least one of those species since then.

If we used tests to measure something about one sex or the other in our own species as you describe, so it only really matters to one sex, then the other sex would still be free to go along for the ride and be affected by the change, or not. In fact, I can give an example of it, or something like it, that doesn't seem to be connected to brain action (at least not very directly). Humans and our ancestors, over the last several million years, have kept getting more and more lightly built, with narrower bones, less muscle mass as a percentage of total mass, and smaller tendon attachment sites on the bones. Given the obvious differences between male & female skeletons & muscles, I'm sure you can see that this means we've been getting more "feminine" (at least as femininity is defined in this lineage; lions would agree, but bears and eagles would not). But it includes the men getting that way over time too!

There are only a few possible explanations. If it benefits females significantly more than males (and might not benefit males at all, as it seems to actually be detrimental), then males are just being dragged in that direction as a side effect. If it benefits both sexes about equally, then some quirk about the biochemical mechanism by which it works just happens to make it affect females more strongly than it affects males, as a side effect.

The question then would be how the mutation we're selecting for does what it does. If it responds to or is influenced by ambient hormone levels, it will be sex-linked. If it's only affected by conditions that are the same in both sexes such as the levels of certain vitamins or minerals and water and so on, then it won't. It could even appear to be sex-linked without the mechanism having anything to do with X or Y chromosomes or hormone levels. If, for example, it gave a performance boost only to people with a high or low basal metabolic rate, then it would primarily affect only one sex or the other because the male basal metabolic rate is higher than the female one (even in childhood).