[SGVLUG] The tangential flogging continues (was: bit order (was Linux based web-server appliance))

[TOM EMERSOM]

> -----Original Message----- Of Michael Proctor-Smith
> On 5/24/06, Dustin Laurence <dustin at laurences.net> wrote: 
> > [...] Do you care to argue 
> > that Intel, which didn't publicly care about power 
> > consumption ... until the P4's started to dissipate ~200W, was 
> > worrying about power consumption in the 486 era? ...
> 
> They actually some what did. First they added halt 
> instruction(something like that) which lowered total power 
> usage when the system is idle. I don't think most OSs 
> supported it or used it.

Ironically, I've heard that the "HALT" instruction, on really old CPU's,
tends to produce MORE heat...

I was going to post a link to a bit of computer lore I once heard, but
now cannot find -- the story goes that someone once executed the "halt"
instruction on a computer using CORE memory [look it up] and the
processor did indeed "halt", however the memory access circuitry was
still running and resulted in the continual reading of the same location
in memory.  Since this is a physical process ["reading" core memory
destroys the contents of the cell, so the circuitry automatically
re-writes the value just read...] that causes lots of "stress" to the
cell [think "flux"] a natural by-product is heat -- eventually the core
memory caught fire...

Most searches for "core memory, halt instruction, fire" returns a list
of "truly useful opcodes", which in turn reference the opcode HCF --
Halt and Catch Fire -- Given the bit of lore above, I'd imagine there is
an element of truth to the origin of that opcode...

The closest I got to this, however, was this page:

http://catless.ncl.ac.uk/Risks/5.6.html#subj2.2

In which someone noted the physical location of core memory closest to
an overtemp sensor and wrote a program to shut off the computer...


But, in the ever obscure way in which google returns search results,
this page was a good trip down memory lane...

http://wwwcsif.cs.ucdavis.edu/~csclub/museum/cpu.html

[and now I'm late for an appointment...]