This is why I mostly disregard early rumors.
A third of the time, I figure it's just over-eager web sites looking for cheap clicks - because everyone hopes a legit scoop is out there, and rumors make for easy click bait. I wouldn't put it above some from just making stuff up, but mostly I think it's people thinking they saw something, or misinterpreting what they may have seen or heard, or just plain wishing out loud and people talking it as some sort of factual release note.
Another third of the time, I figure you do have some legitimate "leaker", but they are going off pre-release engineering samples at best, or just design notes from god-knows-what stage of the design process at worst.
And that final third of the time, I think it's the company itself intentionally leaking information, but doing so on the downlow so they don't have to admit it if they decide to backtrack or if there is blowback. That isn't the same thing as saying that it's legit info - I think they would leak knowingly false information for any number of reasons: to tamp down expectations of their own product, to guage customer interest or disapproval, to attempt to deflect or draw attention away from the competition, etc.
---
I should also clarify, that isn't a perfect 33.3% split between those three options, just I seem to always think the rumors fall into one of those buckets. I couldn't tell you what the ratio is, but I suspect the first and the last a lot more than the middle one. For the particular rumor of nVidia significantly bumping the TDP caps - I strongly suspect it was a trial balloon floated by nVidia to see if people would utterly freak out or just accept it.