Well, in practice, whenever chrome would lock into the 650, the rMBP would heat up to 60-70c just displaying text on a page.
When running Win7 via bootcamp (and vmware, but that was to be expected) the computer is just as hot because it's locked into the 650 - there are no optimus drivers for the rMBP's motherboard on win7 so bootcamp just forces discrete gpu since a major reason for users to install windows is for gaming.
I've managed to, in my brain, distill the likely culprits to the Intel HD side of the gpu conflagration that is the rMBP. Its either something with the integrated gpu or with the switching mechanism that's not playing well with what's written to the shaders by these browsers. I blame the hardware/kernel extensions because the GT650 seems to get along famously with anything you throw at it.
During class now, I'm gonna do a small experiment. I'll delete cache.db and reboot my mac. Then force integrated gpu with gfxCardStatus, start a timer, and run safari. I'll surf until the glitches become "bothersome" which for me is when I see vertical tearing while scrolling (a chunk of the screen doesn't want to scroll and persists while the rest moves) or the flickering text garbage.
Then, I'll do the same procedure again, but this time forcing gt650 also with gfxCardStatus (not syspref for the sake of the experiment). Since I don't expect any glitches to happen during this part, I'll go up to twice as long as the integrated gpu lasted before crapping out.
If neither forced gpu makes glitches, then we can be reasonably sure that its the switching function that "causes" the problem. Whatever makes things easier for the apple devs, right? Maybe then they'll actually fix it.
I'll keep track of my battery drain and temperature during both experiments. This'll be good just to have on the forum for reference, but if more people could replicate, then we could actually analyze the results. Everyone's system is different and you look at different sites so there's a lot of confounds to the data.
