"Yes, genes vary much more within human populations than between them, but these two kinds of genetic variation are not comparable. A population boundary typically coincides with a geographic or ecological barrier, such as a change from one vegetation zone to another or, in humans, a change from one way of life to another. It thus separates not only different populations but also differing pressures of natural selection. This is why genetic variation within a population differs qualitatively from genetic variation between populations. The first kind cannot be ironed out by similar selection pressures and thus tends to involve genes of little or no selective value. The second kind occurs across population boundaries, which tend to separate different ecosystems, different vegetation zones, different ways of life ... and different selection pressures. So the genes matter a lot more.
This isn't just theory. We see the same genetic overlap between many sibling species that are nonetheless distinct anatomically and behaviorally. Because such species have arisen over a relatively short span of time, like human populations, they have been made different primarily by natural selection, so the genetic differences between them are more likely to have adaptive, functional consequences ... as opposed to "junk variability" that slowly accumulates over time."
[Speculation: Perhaps race/population differences in intelligence are caused by a small/narrow set of genes, while individual differences in intelligence ("junk variability") are caused by a huge number of genes with very small effect (per gene).]