autopsy of dead server

justin gedge jgedge at amis.com
Sat Jun 4 18:04:41 MDT 2005


It's been a while since I've torn into PC architectures at this level... 
but here's some info I gathered over the years.

A few years ago -- lightning struck the main office where I was 
working.  All the computers had Tripp-Lite ISO-BAR surge suppressors 
between them and the building power-- some had UPS's attached between 
the ISO-BAR and the computer.  a few computers were still affected.  
There were some SUN boxes [probably Ultra10's] that lost their network 
cards.  The surge supressors protected from HV event  through the power 
supply-- but the network cables running through the building acted like 
huge anteanas to capture all the charge in the air from the lightnin.  
With these computers-- a maintenance call where the network cards were 
replaced solved the problem.  In the case with Sun hardware-- the 
HV/lightning damage was limitted to the network cards.  I'm not sure how 
PC's go in this area-- but if you didn't have a huge storm-- them this 
likely not the cause.

As far as power supply causing trouble-- I've got some info on this.  If 
everything is fried -- but your processor still works-- it's probable 
that something happened to the system power.  I've seen cases where some 
amateure hooks up power backwards [reverse power to the motherboard] and 
everything on the board was smoked out except the processor [memory 
might be salvagable also].  Reason:  typically your computer power 
supply has +12V, +5V, -5V supply.  Back in the early `90's as cmos 
geometries got smaller and smaller the supply power had to go down [for 
the uP].  Partly due to physical breakdown of cmos transistors at 
smaller feature sizes and partly because lowering the voltage helps with 
reducing the power consumed [and heat dissipation through the chip].  To 
accomodate these new processors that were at 3.3V and lower-- the 
motherboards needed a new chip to regulate power for the uP and other 
low voltage circuitry.  Typically a drop out, or low drop out [LDO] 
voltage regulator is used.  Most LDO's [and voltage regulators in 
general] are built on BJT [Bipolar technology NPN/PNP transistors] 
instead of cmos technology.  By it's very nature bipolar technology is 
more immune to ESD [Electro Static Discharge].  The LDO in many cases 
will tollerate ESD events better than a uP and most are designed to 
protect downline power from reverse battery- power spikes, esd, etc... 
so if your uP is good -- and everythign else went south on the server -- 
it's a good sign that your power supply went out-- or some other power 
related event took the system out.


Justin Gedge




Jordan Curzon wrote:

>I have a computer that simply died. The motherboard won't boot and the
>two NICs that were on it give PCI errors in other boards. Any ideas as
>to how to determine post-mortem, what killed it?
>
>Jordan Curzon
>.=================================.
>| This has been a P.L.U.G. mailing. |
>|      Don't Fear the Penguin.      |
>|  IRC: #utah at irc.freenode.net   |
>`================================='
>  
>




More information about the PLUG mailing list