This group is so dead I thought I'd share a saga about my old hardware\nreliable 503+ system.\n\nI had some problems with my 503+ system that uses a k6-3+ at 550Mhz. I\nthought it was the PS since the PS fan was locked up, so swapped in\nanother PS, which in fact was a better quality supply made by Delta.\nOddly, it was several days after installing the new PS that the system\ncrashed/locked, I reset it and walked away to the next room over to\nwork at another computer. The printer connected to the 503+ was not\naccessible over the network, so I walked back into the room where the\n503+ system lives, and omg, the smell of something burning permeated\nthe air.\n\nI killed power and opened the case and did a sniff test, found the CPU\nto smell the worst, so I touched the HSF, and it was hotter than a\nfirecracker. Further analysis revealed the Vcore PS circuit had 2\nbulging electrolytic caps. OH NO me sez, thinking the CPU has just\nfried. I cycled power just long enuf to get a BIOS boot screen, and\nto my amazement, the system tried to run. I quickly power down, and\npulled the CPU, and started checking some voltages. Seems the CPU was\ngetting 3.8v, that's right, 3.8v, only 1.6 volts above the max spec\nfor the CPU. I surmise that the bad fan in the old PS may have\ncaused overheating in the box, and closer inspection revealed the\nVcore switch mode power supply circuit was having problems.\n\nI did a little circuit probing, and replaced the swollen caps, and a\nvoltage test revealed the CPU socket was still getting 3.8v still (had\nthe CPU out of the socket during these tests just in case). The\nsignal from the PWM controller should have the MOSFET turned off, but\nit was conducting lots of current to the CPU socket. Sure looked like\nthe MOSFET in the Vcore circuit had died (it's drain-source impedance\nwas quite low with no power on the system). I searched the net for\ninfo on the device that failed, it is an n-channel enhanced mode\nMOSFET made by Hitachi. I found the data sheet on the 8-pin Vcore PWM\ncontroller chip (it sits right by the MOSFET device), and a typical\ncircuit schematic was provided by the manufacturer in its data sheet\ndetailing how such a CPU PS circuit should be built using a MOSFET\ndevice. A brief survey of some old boards I keep around to cabbage\nparts out of yielded a good quality Philips MOSFET device of similar\nspecs (I have to complement Abit in their AX7 design, I've stolen all\nthe PS caps out of this old socket7 board, and now I'm stealing the\nCPU PS components and all were good name brand stuff). Philips has a\nnice web site for obtaining data sheets on their semiconductor\ncomponents, I tip my hat to their IT department.\n\nThe worst part was getting the old MOSFET off the circuit board, some\ntrace damage occurred since the tab/body of the MOSFET device is\nsoldered to a large area copper trace on the mobo, and it's darn\ndifficult to get a enuf heat into such areas to reflow the solder and\nseparate the device from the mobo without special tools. Fortunately,\nenuf of the copper pad remained (after I tore most of the copper off\nbeneath the MOSFET) to build up some solder onto for soldering to the\nreplacement MOSFET body tab, which was not actually designed for\nsurface mount, but bah who cares, a little jerry-rigging, lead\nbending, and solder built up on the tab of the replacement device and\nI had it attached in place. With fingers crossed, I hit the power\nswitch (had pulled all boards out of the mobo, and pulled the CPU) and\nall was right with the world again as I measured 2.1v in the CPU\nsocket (I use 2.1v with my k63+, seems to be what it likes). I then\ndeveloped a big smile on my face cuz things are back where they should\nbe, then I drop the CPU and the boards back in, hook up all the\ncables, and the old workhorse is back alive again.\n\nI do recall a dead HS fan on an old AMD233 k6 system where I burned HS\nfin stripes in my finger when I touched the HS, and it appears a k63+\ncan also take some serious overvoltage and heat abuse without failing.\nIt's so satisfying to fix a problem like this w/o even buying a single\ncomponent!