Theoretically, it does not make a difference for SEO unless the code contains some very egregious coding errors. Matt Cutts has said that most sites do not pass validation and therefore they do not use that as a factor. Google's simple home page does not pass the W3C validation. It shows over 40 coding errors.
I have seen situations where a search engine was not spidering a site and my investigation found more than 250 coding errors on each page. Fixing those issues did resolve the problem, but that was because a few of the errors were what I would call egregious errors due to code that was horribly structured.
Where W3C validation does help is with site performance and consistent rendering across different browsers. This doesn't usually help with CSS browser incompatibility issues, but is does help with the rendering of the general structure of a page.
Beyond that, you can still have a web page that has horrible structure to the point of being unreadable by search engine spiders, yet it will still pass the W3C validation. I have seen that many times when some designer using WYSIWYG tools designs a web page with a complex table structure that has 5 to 10 nested tables and the content is buried in the middle. You end up with a script that is about 95% code and 5% content, which is a bad mix. When you do something like that, the spiders may not find the content.
IMHO, W3C validation is a good tool for identifying potential rendering issues, and a good way to learn about coding standards, but it will not have a significant impact on your SEO results. Nonetheless, I do frequently use the validation tool just to make sure that the code meets the standards.
"Democracy is two wolves and a lamb voting on what to have for lunch. Liberty is a well-armed lamb contesting the vote." -- Benjamin Franklin