Re: Physics - was Re: Mem upgrade for 701C

[jiml@strauss.ncrmicro.ncr.com (Jim Lewczyk)]

My two cents, 

> >OK, I'll try once more and then I give up ;-)  .
> >
> >E = I/R
> yes, i understand your problem.  i don't mean to change this list into a
> physics lecture, so unless i have this wrong, let this be the end.
>
>At the risk of being shot ...

Ok, ok, I don't want to prolong this either, but I work for a semiconductor 
manufacturer, and I have a bit of perspective.

We make several parts that are both 5V and 3.3V qualified.  You can plug into
either supply and get the function you expect.  Intel's DX2-66 CPU is a 5V/3.3V
part as well (AMD's is not).  It's a matter of operating regions and noise 
margins within the semiconductor.

Semiconductors are not linear devices when it comes to current draw, but they're
not constant-current devices either (usually).  A 5V memory module will probably
draw more current at 3.3V, but the total wattage consumed will probably be less
overall.  If it works at all.

Clocked parts are a different situation.  It's possible for a clocked part
to latch-up if the voltage is wrong, and if it latches up in a high-current
state (that would normally be transient), it could cause overcurrent problems 
(usually, however, high current draw comes from a fast clock).  An odd side 
effect of the lower voltage is that the speed rating of the part may actually
go up.  I don't have any kind of intutive feel for this, that a 70nS part will
operate at 69.5nS or 60nS, but an asynchronous part (ie, unclocked) requires
setup time to change states.  If the state or voltage it reaches is 3V away
instead of 5V, then it'll get there sooner and settle sooner.

If you plug a 5V-only DIMM into a 3.3V socket, it will either work or it won't.  
It won't "burn out" the socket.  My biggest concern, assuming it worked, would 
be if the margin of operation of the chip stayed reasonable.  I wouldn't be too 
happy to have the thing flake out on me after an hour in the summertime because 
it got too hot, or stopped working when the battery got low.

If someone has purchased the 5V $285 16Mb module and has run successfully at 
the hot corner of the operable range (high temperature, say, outside in Texas 
or New Mexico on a 120 degree day, low battery power, and a long time run) with 
no problems, I'd feel better about it.  However, unless the seller guarantees 
that the part is really qualified at 3.3V you're taking a chance to save $144.

The last ad I read from Joseph Gadoury (jgad@earthlink.net) he WAS offerring
a 1 year warranty.  Now the question is, will he be around to honor it?

Personally, I'll take the chance on the part, but I have to fix my laptop
first ;^)

The IBM part isn't proprietary.  I've put a 4Mb module from a Dell Latitude 
XP into my Butterfly and it was a perfect fit (but I didn't turn it on.  My 
Butterfly has other problems at the moment and I wouldn't have been able to 
tell if the machine even saw it).

I'm glad to see the trend in standard memory parts in laptops.  All that 
remains is for the manufacturers to stop pretending that they're still
proprietary.

Jim Lewczyk