Electronics

So I was on the EVGA forums as this guy has 2 dead 7900 GTX cards that he wants to give away. I say "Sure, why not!" and aobut 3 weeks later I have them sitting on my desk! All the while I've been researching how to repair electronics to give me ideas on how to make these work. The #1 problem is typically blown capacitors which are easy to replace. Well, it's gonna be harder than that with these cards (I think the guy my know more about fixing them than I do). There are a total of 4 capacitors on the card and all look to be in good shape. So here's a picture of the cards:


As you can see they only have the heatsink on the RAM. He won an Alienware system and it came with water cooling so the stock cooler didn't come with them. So in order to test them out I had to put my cooler on them temporarily (I have a 7900 GT). Initial results show that one card may have bad RAM, not easy to fix. Basically the card had a hard time loading the test and paused between loading scenes which is what gives the hint that RAM is the issue. The other card started out great but slowly degraded over the testing, heat was a big factor there. Either the cooling wasn't sufficient (not as likely) or the voltage regulation is wacked. I will be doing further tests with that card once I can find a cooler for it. If anyone has one lying around I'll be happy to take it off you hands ;). The cooler for the GTX is quite a bit different from the GT that I have; both cover the GPU and RAM but the GTX also covers cooling of the MOFSET; that black heatsink that you see on the right of the card (which go really hot with both cards).

One thought with the card that worked at first is the RAM heatsink that is on it (the large copper thing on the board). If you look on the right side you will notice a black square that touches the heatsink just before it takes that 45 degree turn. That is a voltage regulator similar to the MOFSET under the black heatsink. Well, it does touch the heatsink, infact the heatsink is putting a little bit of pressure on it. It makes me wonder if that regulator went bad from the added stress, of course I could be wrong.

Also there are thermal pads between the heatsink and the RAM chips. Whoever it was at Alienware that put this together didn't pay close attention because most of the chips were not completely covered by the thermal tape. This was the case with both cards. Uneven temperatures across a chip can't be good for it. From what I understand about how RAM works it cycles power through it to keep the memory state. Electricity will flow through the chip which generates heat and temperature affects how fast the electricity moves; cold = slower, hot = faster. If this is right (again I'm no expert) then that would put additional stress on the chip causing it to wear faster. This may be the cause of RAM failure in the other card, of course it could simply be a bad solder joint too but solder joints on these chips are a ball joint that is heated and sandwiched for the connection and requires special tools to solder/re-solder.

Ok, Ok, enough babbling about electronics, I like challenges and this is looking to be a pretty good one. Hopefully I can get at least one card working :)