I’m rather sure most of you won’t believe this without having seen it, but I am not the first to try something absolutely ridiculous to handle such problems. In fact I got the idea from a video I saw some time ago, might have been on hackaday even. The HP TX 1000 has a common problem with its GPU that gets way too hot and sooner or later this leads to a black screen for many owners. One guy has fixed this with household items and I was quite impressed when I saw him making a heat shield out of cardboard and aluminum to keep the rest of the motherboard safe from his soldering equipment, which was, iirc, a 75W reflector light bulb. I mean the kind that looks like a mushroom if you aim aim it upwards, this here:
He just put the thing right onto the GPU, waited a certain time (I don’t remember how he came to the specific time) and claimed later that after this the thing would be working again. That’s the kind of thing that really makes you wonder if it is a bad joke, coincidence or actually working if you have ever read how complicated to handle reflow soldering is said to be. I later came across several descriptions for repairing some board in the old CRT based iMacs that was often making problems. These involved packing the thing in aluminum foil for shielding and then baking it in the oven, like a damn turkey! And there were quite a lot of people claiming that this has worked for them.
So quite some time ago I got a slightly defective Acer 2001WLMi. I was told it doesn’t boot and it has formerly been crashing or freezing randomly. I agreed to trade the hardware for a backup of the remaining data on its disk. To my surprise it booted and worked seemingly flawlessly, but I later experienced the described random crashes. After some hours it just froze, sometimes even after some minutes. Partly disassembling it and trying to stress it in various ways made me suspect the graphics card as the cause of the problems. Stressing it with some 3D benchmarks seemed to made the crashes appear faster and suddenly most of the time there were randomly colored pixels present when the screen froze. Cooling it didn’t help that, but physically pressing on the card on random places made it fail pretty easily. Too easily for my judgment.
So after I was out of ideas and had already broken the display cable by repeatedly taking it in and out I decided to challenge my luck with trying to resolder the memory and the GPU. I could lose a graphics card and a mainboard not really working and not really worth anything or win by making it work again. So why not give it a shot? After thinking what is lying around in the house, it was Saturday and past midnight, so, er, technically Sunday, that might be fit for the job I thought of the 500W light in the basement and fetched the aluminum foil from the kitchen.
I’ve been told people go crazy for random useless images between text and this one actually looks exactly like the lamp I used. To show off bad handwriting and a lack of drawing skills I even made a rather useless drawing of the setup. The wood blocks were roughly half an inch in height and using more space fails to produce the needed temperatures.
I put small chunks of solder on the aluminum foil around the chips to have a rough indication of the temperatures present there. Also I put the temperature probe of my multimeter on one of the chip’s surface. After all the solder is on their underside. It turned out the temperature rose very fast to something like 100 degrees, but from there on it took some time to reach something like 140. I decided it’s a good thing if the temp rises slowly, too fast is indeed bad, and played a bit of Minesweeper while having an occasional look on the thing via the temperature probe and a mirror and sunglasses for the soldering chunks. When the upside of the chips were around 170 the solder chunks had turned into balls. Based on some reflow soldering profile I had read I guesstimated that I might leave it heating something like maybe 4 minutes.
To make the cooldown not too fast I added a second wood block on the side where the chips were and immediately temperatures fell again. After two minutes I added more height and left it for another 10 minutes before switching the lamp off. To my surprise the bloody thing worked still – but kept crashing. So I did the exact same thing with the GPU itself. And guess what, the thing is now working again. At least I had no more crashes when pushing around on the PCB and having it run 3D benchmarks over night.
Well, since then the laptop has been lying around on a shelve and to this day I haven’t found an acceptable offer for a display cable. It’s good to know that I have an oversized thin client lying around, though. One that’s even capable of Windows Aero. It would even make a bearable office PC still, with a Pentium M 1.6 and the two GB of DDR RAM I have in stash. If I don’t find a proper cable this thing’s going to the guest room in the basement, where the emergency screen, a 21″ CRt is also lying around…