20121201 old computers sometimes take too long to di - plembo/onemoretech GitHub Wiki

title: Old computers sometimes take too long to die link: https://onemoretech.wordpress.com/2012/12/01/old-computers-sometimes-take-too-long-to-di/ author: lembobro description: post_id: 3768 created: 2012/12/01 23:44:16 created_gmt: 2012/12/02 03:44:16 comment_status: closed post_name: old-computers-sometimes-take-too-long-to-di status: publish post_type: post

Old computers sometimes take too long to die

Just pulled off another rescue of the extra Dell Dimension 5150 that we've been using as a Tekkit server/media station here at Casa Lembo for the last few months, and am now experiencing the inevitable disappointment at again missing the chance to put it out to the curb once and for all. The machine in question was one of a pair we got in the summer of 2006 just after moving down to NC. The other one was actually one of Dell's "open source" machines preloaded with FreeDOS instead of Windows, and still serves as my main Linux workstation at home. Starting a few months ago its sibling got reimaged with Scientific Linux 6 for use as a media server. Later on Tekkit got installed -- just for the fun of it. The trouble started when we had a power outtage one night at the beginning of this week. When the lights came back on this computer didn't. After some snooping around I diagnosed the problem as due to a power supply failure. It's actually pretty common for machines over 5 years old, and given that the unit was plugged directly into the wall (and not a surge suppressor) it made perfect sense in this case. One test I performed involved jumping the supply at pin 16 and 17 (green and black wires, respectively) to see if I could power it up. This worked, but only if the DVD player wasn't powered. That told me there were problems in both the internal switch and at least one of the leads going out to the DVD. In any case not something that could be fixed. The supply needed to be replaced. Fortunately the 5150 can use a standard ATX power supply. All of the power connections and pinouts are just what you'd find in any generic system with an ATX motherboard. It only took a few minutes to drive down to the local computer store and pick up a substitute unit, the Cooler Master Elite 350W PS (the original was a 305W Dell W8185/L305N-00). Because the SATA device power connectors on the original supply had angled connectors I also bought a Molex to angle SATA adapter. The new unit fit perfectly, with a little nudging to get the retaining screws lined up. The ATX and Aux power connectors fit into their sockets in the motherboard as expected, as did the Molex connectors that went out to the DVD and SATA drives. I had to work a bit to make sure the wire bundle for the replacement wouldn't break up the airflow inside the case (the old unit's wires were "just long enough", while wire lengths for the new one were generous enough to support a full tower configuration). After buttoning everything up the machine came right to life when I plugged it in (the BIOS was set to start on A/C restore). One of the great advantages of early PC's was the fact that they used off-the-shelf, interchangeable, components. All of the major hardware vendors violated the spirit of this design at one time or another, the more eggregious being IBM's abortive effort to corner the market through its proprietary PS/2 architecture. Whether it is a low profile "all-in-one" or a standard desktop, most big vendors still seem to think that highly customized components give them some kind of market advantage (customer lock in?), something I really wish someone would disabuse them of. I can say with confidence that I couldn't have used an off-the-shelf part to make the same repair on many models of Lenovos, HPs, Dells, or Gateways, to name a few. Of course the real problem is that I succeeded, thereby delaying a long wished-for purchase of new hardware that can support much needed requirments like 8 or more gigabytes of RAM for gluttonous Java server software. For now we'll have to continue to work within near crippling limitations like a maximum accessible memory limit of 3.5 Gb (not to mention a "64-bit" CPU architecture that's missing the necessary extensions to run 64-bit virtual machines...). And weep.

Copyright 2004-2019 Phil Lembo