Thanks Kerry. I've thought about getting a ups, but since I'm not an isp with thousands of users, I didn't feel that the $60 or so was justified. Besides, the last significantly long power outage in this area was in late 2000. One of the nice things about the old AT design was indeed the power switch that wasn't controlled by the board. I've always suspected that the battery responsible for running the hardware clock was also responsible for providing power to whatever chip storred the power state. However, I'm speculating here as well. If I'm not here, and I know the power has gone off and come back on, and I'm in front of a terminal on the net, there is always wake on lan I guess. Greg On Wed, Jun 05, 2002 at 04:27:25PM +0800, Kerry Hoath wrote: > This feature is not fail safe. Since the switch > is a software construct then your power supply needs to > provide sufficient standby current in order to keep the board happy. > Old computers had a dual pole single throw switch so when power was on > it stayed on so as soon as the power > came on the whole machine got power. > With the modern ATX design it is up to logic on the board and in the supply to > sense when a key is hit or network is accessed or > modem rings or button is hit and power up the machine. > Depending on how the power goes out; depends on what state the supply > is left in and how the cmos is notified that power has gone if at all. > I have seen bioses claim that they will power up a machine when > power returns; but rarely has this worked all of the time. > Usually the feature works when you don't need it to and fails silently > when you do need it. > I presume some standby current is stored in some capacitors on the board in order to save > some state info in cmos or in a circuit to tell the machine what > the previous state was. (the above sentence is speculation so > corrections but no flames accepted). > If this info is not set correctly or not stored or doesn't work for reasons of the > supply or board or mysterious circumstances the machine > won't power on. > It is possible to get this feature to work some of the time but not reliably. > get yourself a Jell-cell UPS which should > keep the box going for 30-45 minutes assuming no monitor. > You also get controlled shutdown via serial port notification; > and protection against short outages. If you do have an outage long enough to > exhaust the ups then your power on after wibble feature might save you > but don't count on it. > If it is important for a server to come back on and the > server is ATX; I get an UPS or call someone to press > a button. > > Note also that if an ATX supply has power go on and off multiple times rapidly or > has power dips the surge protector can't account for; > it can lock up in such a way that requires you to > turn off the power at the wall for 30 seconds or more to reset the caps in the > supply. I don't know why this occurrs in atx supplies, I think > the switch-mode gets confused. > > Regards, Kerry.
