Some time ago I decided to upgrade my workstation from 512MB RAM to 1.5GB by adding a 1GB DIMM to the already installed 2x256MB DIMMs. I bought and installed it and was pleased with the result - but not very long. The system started to get a hiccup. It got worse and I decided to reinstall my system. Only after about three weeks after doing that, I got BIOS checksum errors and I realized that there must be something wrong with the additional RAM, and fired up Memtest86 which reported errors after a short time.
While swapping the DIMMs around to find out which one
was the bad one I noticed a "dead electronics smell".
Afer some careful investigation of the mainboard I saw the following:
You can see that the small MOSFET below the two caps (the one labelled 20N03L) has got a crack. This one is definitely dead. Additionally, the bigger one in the upper right corner (labelled 55N03LTA) has a sort of brown crust around it, which looks like solder flux. This is typical for overheated parts.
I traced a PCB wire of the small one and ended at an IC which is labelled 2995N and is made by National Semiconductor. The full part number is LP2995 and is a "DDR Termination Regulator". According to the datasheet this is a linear regulator and the MOSFET should not be too critical as long as it has sufficient voltage, current and power ratings. I decided to replace it with a much "bigger" 35N05 (if I remember correctly).
I decided to unsolder the bigger MOSFET too and yes, it was fried too. I thought about the function of it on the mainboard and developed the following theory: There are chances that the RAM power supply couldn't cope with the increased load by the additional DIMM, got hot and blew. This may have supplied a too big voltage to the RAM which in turn would output a too big voltage on the data lines which yould have caused the smaller termination regulator to blow. If this was the case, than this MOSFET was the DIMM regulator. I didn't have a 55A MOSFET for exchange around. As you can see in the picture, there are no coils around the regulator. This lead me to the conclusion that it must be a linear regulator. But if its generating the 1.8V (or so) DDR supply out of the 3.3V PC power supply voltage, the current consumption can't be more than one ampere or so or it would die immediately because of the small (or better: non existant) heatsink. Therefore I decided to put in another 35N05.
I mounted the new-old MOSFETS on nice heatsinks to reduce the risk of
repeating the semiconductor:
I inserted one DIMM and powered the PC up. And - what a wonder - it seemed to work again. I didn't think that my chances were good, because I thought that the failing voltage regulator would probably have fried something else too, but it seems like this wasn't the case. I ran Memtest86 to strain it a bit but it seemed (and still seems) to work correctly. I've only inserted the 1GB DIMM because with all 3 I've got errors. (I must mention here, that I've haad problem with all together from the very beginning, before the MOSFETS decided to smoke. The old ones work flawlessly and the new one too, but they simply don't with each other.) With one DIMM the termination regulator MOSFET doesn't get warm at all and the supply regulator may be about 40°C. No miracle that it dies with 3 DIMMs and the tiny copper area on the mainboard. I consider this as the manufacturer's design flaw.
Usually one has a silent voice in the back of your mind which tries to tell you that you can't repair computer components, because they are so incredibly complicated. It might be true that they are complicated but they are still "simply electronics". If you have the skills, tools and time to repair them it may be worth a try.
Copyright (C) 2008 by Wiesner Thomas
Last change: May 21st 2008