Originally Posted by Shredicus
Actually not quite. Weird story. With my second 570 I was benching it on the new furmark when the latest version was first released and smelled the melting smell, then I shut the PC off at the PSU. Weird thing is it booted up ok and ran fine for about a week. No artifacting or anything, no more burning smell. I gamed with it at its normal overclock repeatedly. The card finally kicked the bucket after a week when I was playing Metro. I took the cooler off and sure enough, melted VRM. The explanation for this from what I read by local GPU wizard pizzaman is the reason these cards are dying from the VRM phases melting is because the metal liquifies and then actually leaks onto the PCB and contacts another elctrical surface causing a short.
I'm not sure your wizard pizzaman knows that to "liquify" the metal in the caps require a bit more than 100c, and he is wrong by a few hundreds of degrees there.
Actually what happens is that the cap of the VRMs melt (they are made of cheap plastic along with the PCB), and the VRM internals does a little "kaboom" from excessive volts and heat. There is no metal "liquifying" of any sorts.
Because of plastic melting, there might be a contact exposure between the layers, yes.
Along with that, the soldering, which is the weakest link, might have answered the question "which is the coolers card around?" wrong and got kicked out by anne robinson, and the VRM is now off from its place and is just being held by the cooler only.
And to the OP:
Search for any burn marks, anything which might look like a melt soldering, or the sort.
I don't know how galaxy warranty, so don't know if its ok to open the card or not.