EDIT:

In all my time of growing up I never heard of anything that made more sense, but god damn seemed so insane... IT WORKED! I'M TICKLED HAPPY!!! I can't believe baking my GTX 470 WORKED!!!!

My procedure:
1. Disassemble down to the card itself.
2. Remove excess thermal paste and make sure to clean the GPU core completely.
3. Preheat your conventional oven to 385.
4. On a baking sheet with aluminum foil underneath, place the card upside down so the back of the card faces up, having it supported by little balls of foil.
5. Cook that bad boy for 10 soild minutes (You'll **** yourself the entire time but I guess its safe :D)
6. Take it out and let it rest, like that roast you made last night, for 20 minutes.
7. Reassemble and test!

I'm still just.... Blown away!!!!!

Ide like to toss an uber thanks to Sajin for the tip, Ide be sitting with a paper weight without you bro. *BRO FIST*

Are you for real? I mean yeah, not like I got anything to lose but a hunk of un-refundable, un-rma-able $400 paperweight, but lol. Sounds like by baking it you're reflowing it. Hell, why not. I'm gonna wait for a second opinion though, I wanna make sure it's a last option hehehe.

epicflux

Are you for real? I mean yeah, not like I got anything to lose but a hunk of un-refundable, un-rma-able $400 paperweight, but lol. Sounds like by baking it you're reflowing it. Hell, why not. I'm gonna wait for a second opinion though, I wanna make sure it's a last option hehehe.

 

No Problem, If you end up cooking it make sure to post back with the results.

Whoa!!! That's insane!!! I wouldn't do that!!! It's too scary!!! Glad for you :D

I am glad this worked for you, but your "fix" comes with a major caveat.  The reason that this works is because you are, in essence, melting the outer membrane of each solder joint.  You are not, I repeat, NOT, refreshing the solder joints, even though it appears so.  An oven is defintely not hot enough to accomplish this task, and it will absolutely shorten the overall lifespan of the card.  The reason it shortens the lifespan is because each time this band-aid fix is used, the solder joints crystalize.  The more they crystalize, the higher the temperature is necessary to resolder the joint.  The result is a card that can no longer be properly repaired.  To properly repair solder joints, one needs a BGA welder rig (which uses infra-red radiation) to heat solder joints to extreme temperatures, and, because of the nature of infra-red heat, saturates each joint with head, inside and out, resulting in a complete remelt of the joint, which leaves no crystalization.  If the joint isn't completely resoldered, it will fail again, and each time it fails, it will take less and less stress for it to fail again.
 
My intent with this post is not to spoil anyone's fix, but to allow them to make an educated decision regarding said fix.  For those who need their cards repaired in a proper way, seek out a repair shop that uses a BGA welder (a sure indication that they know what they are doing, considering the cost of said welder (well into the thousands US$).

Well it wasn't permanent. It froze today and started artifacting again but this time it was a little different. I felt that card and it was damn hot. I think it's because I didn't apply fresh thermal paste (their was so much I just scraped off what was on the core and reused what was on the heatsink) so I'm doing it again now, ultra cleaned it and applying new thermal everything :D

If it doesn't last this time then I'm just gonna have to step down and put out more doe for another card. Either way to see this is still somewhat viable and possible I'm happy. And I'm not to worried. Plenty of reports online say their cards still last up to a year with this method with a few periodic rebaking which I'm doing more prepared this time.

As for the results on the first round, everything was good except for when I ran the UDK NightandDay test map, whenever it going to the second scene where it has a focused close up demonstrating blur effects the textures appear to flicker black and normal. Strange, other then that, I run Unigine Heaven maxed extreme tesselation and the results end up like I usually get while also being a little higher thanks to the new beta drivers :D

I'm sorry I didn't mean it as in it's all me. I guess that was heavily miss-worded. If its you who originally came up with the idea than kudos to you :D

Btw rebaked, added new thermal paste (Dynex brand... yay...), and now It's running perfect again. Stress tested on Crysis for a bit. Works great so far. Lets see if I can get it to atleast last a month :D I'll keep you all posted. Spread the word! 
 
When I said "My Procedure" I meant what I did. Guess I should have it say The Procedure, or if you came up with it, HeavyHemi's Procedure 

very true and does work. only use this solder re flow method on older gpus out of warranty.

Ok well I left it on all night to check for stability and now it's doing it AGAIN!!! I don't know if I'm not making the oven hot enough or leaving it long enough but it appears to be only a one night stand each time. 

My theory, is that the GF100 is just too hot. These line of GFX cards run in the 90s+ in games so I'm guessing this micro fissure is not really being repaired and when it starts to run hot for a little bit, it causes the problem again. Just not that viable for as GTX 400/500 series card I guess.

My final attempt will be tonight at 410 degrees for 13 minutes. Quite frankly maybe I'll just let it melt to nothing, take a couple photos, then trash it. Unless someone wants to buy it and maybe test out some methods to fix it themselves or someone who can reball these things would probably be in luck. Ide sell it for $100 considering I payed $400 and it's still in a possibility to be fixed considering it is a solder issue.

If their was a way to isolate where the issue was at ide resolder it myself :/