A quick update: all artifacts have disappeared and the display is rock solid. This is without another reflow and also without stress testing. Temperatures as reported by the sensor seem nominal at about 35c. The cpu/gpu fan runs quite a bit faster than it did when the computer was "broke" so I'm still going to disassemble it to make sure I didn't leave a sensor unplugged and also check the heatsink-gpu surfaces.
And btw, I'm guessing there's no solder inside the die so how does heat repair the connections even temporarily ?
The problem might not be the bga under the chip. The crystal is also connected to the die via solder balls. They are a lot smaller, but still solder.
Also you can take a look at this, it should clarify why heat effects it. https://books.google.no/books?id=G9A...%20die&f=false
Also if you had your station set at 450F the first time, that would explain why there still are artifacts. The air gets a lot cooler by the time it gets to the chip, and if the problem is the solder under the chip it's even cooler when it reaches it. I wouldn't be surprised if the chip was at under 150C (302F) during the reflow. But this is good! If the chip starts to show signs of life with only a low temp reflow, that means that there is a good chance for it to be fixed by a proper one.
I am talking Celsius. Really. Remember the hot air station will have its sensor inside, in front of the heater element. By the time the air goes thru the nozzle and into the outside world, it cools down significantly.
If you use a multimeter probe put in front of the nozzle of your hot air gun, it should read around 320C. I can't see how lead-free solder would melt any lower than that anyway. I use 400C for working with leaded solder and 430-450 for lead-free. I don't think my station is poorly calibrated as this is my 2nd one and i'm using the exact same settings. My hot air station is a ZD-939B.
I have however found that for Gordak ones you need to set the temp lower, as the heater seems to be higher power. The Gordak i had at work did not have its heater marked in deg C, just from 1 to 8. I had to use it at the "5" position or slightly lower on sensitive boards otherwise it would burn them.
There is solder inside the die, obviously. How do you think they connect them otherwise? They don't use bond wires anymore for chips this dense.
The VRAM does not get hot, there's no problem leaving the old pads on.
Oh, so you were talking deg F. I was talking Celsius. Does it even melt solder at 450F? You will want to use a setting that you would normally use for desoldering small parts.
Remember to reapply good quality thermal compound. And don't be afraid to blast the die with really hot air - just don't do the same for the green substrate.
Ya I hit it at 450f but I only ever use the station at 500f as it is mostly useless below it but I figured after four minutes it would work. I've got no problems hitting it again at 500
I will say I only have generic paste but have never had a problem with it the several times I've used it, lastly on an LG z-sustain. How hot does the vram get ? I've been reusing whatever goop is on there already as I have no thermal pads let alone know what thickness to use.
Remember to reapply good quality thermal compound. And don't be afraid to blast the die with really hot air - just don't do the same for the green substrate.
Reflow did not help - symptoms remain unchanged. Though I'm not averse to doing this several times for the learning aspect (computer doesn't work anyway!) I'm loathe to keep disassembling this thing for a temporary fix. I'm still trying to find the type of ram used on the listing I posted above before I make my next move.
Okay, that is a desktop chip. It happens to be exactly the one i had linked in the ebay auction above. So if you break it, you know what you need.
The bumps are in the die, the black square. Concentrate all your efforts on there. Heat it for a good 3-4 minutes with the air set to around 10 o'clock and temperature to what you would normally set it to for, say, soldering a QFN chip with lead-free solder. In my case this is around 450C on the dial.
Have a heatsink on hand or something else that you can apply some pressure with without burning yourself. When you lift the wand away from the die, apply slight pressure to the die and hold it there for a minute or so. Then re-install the card and test it. Should work, but remember it is a temporary fix, and how long it lasts depends mostly on luck.
Avoid heating the substrate too much or it will bloat and short out the chip. Also known as "popcorning".
Got it. Tore it down again and you were right. And on top of it the date code is 0632 not 0832 so this is probably original.
I hit it with some heat but either this is too far gone or it wasn't enough heat (my wand can't effectively wet large ics). Reassembled and the symptoms had changed just not cured.
Question on these bumps.. are they in the die or throughout the substrate as well ? Just for fun I'm going to try and hit it again and would like to know where to concentrate my efforts.
If it says GO7300 instead of G73 on the chip then it is the mobile version optimized for lower power consumption (and thus lower performing). The G73 would be what you would find in a desktop pci-e 7300GT card. They are different chips and one cannot replace the other. Just pointing that out in case you want to go the route of replacing the BGA on your existing card.
GO7300 is a different animal. It is the mobile (laptop) version. Not surprising to see the GO7300 on it since it is a MXM card. And there are several revisions of that, all compatible with each other tho. But they come in two different physical sizes so care must be taken. Also, it is a high demand chip so it's a bit more expensive, it runs about $35.
I agree with you on getting a 7600 rather than getting this 7300 repaired. Besides being more powerful, there are plenty of them which still live to this day despite running hot, so with good thermal paste it will last you a good while. Seems like it was built a little bit better than the other chips affected by Bumpgate.
Check this out http://forums.macrumors.com/showthread.php?t=877441. They mention a tool that can generate the BIOS, but I'm not sure if it can generate a 7300GT BIOS.
If you can't find a BIOS file, can't you just dump it from the old card?
Leave a comment: