PDA

View Full Version : Had this wild idea for cpu placement, but also some concerns...


=ACID RAIN=
01-31-04, 03:57 PM
I was thinking about the socket platform, and started thinking about how the CPU could be modded for better cooling (going to use socket-7 as an example, since I do have some socket-7 parts laying around). Somewhere along this thought I came up with a strange idea - why not mount the cpu in another location, like the top of a case, and run wires from the second (external) socket down to the motherboard socket? I thought about that for a moment and came up with a few potential problems:

-Crosstalk
-physical distance = performance hit
-where to mount the cpu socket?
-what wire layout (combined with crosstalk issue)
-motherboard socket interface

So let's pretend for a moment that we don't care about a performance hit. That leaves a gaping concern for crosstalk. To fix that, maybe a series of ribbon (IDE or SCSI) cables could be used, with every other conductor being grounded, so crosstalk (and maybe some external noise) would hopefully be eliminated.

Ok, so with noise "hopefully" taken care of and wire layout out of the way, the problem comes to how I will connect the cpu to the motherboard. Two methods come to mind:

-solder wires to the contact points on the cpu (pain in the ass)
-find an adapter of some sort with an easier solution

I saw a socket-7 to socket-5 adapter at work somewhere, so I'll look for it again, it maybe easier to solder on than a plain cpu. From there a cpu can be dropped in hopefully.

At the other end of the cables, how could one interface a socket on the motherboard without sticking wire-after-wire into the socket holes, one at a time? One thing I thought of here would be to take a winchip, which normally has a green PCB under the cpu (and is easily removed), and make a plug-in module for the socket. This would greatly simplify trying to get the cpu interfaced properly.

I'm open to ideas, and if this has been done before, I'd like to see a link :). If this has never worked, but has been tried, let me know so I don't waste my time and effort on it without results plz!

pranks7er
01-31-04, 06:24 PM
interesting idea try it out on olddddd processors first

DaveSauce
01-31-04, 06:44 PM
hmmmm that's a definate no right there......

motherboards are SENSITIVE circuitry...all the traces are carefully planned out, and all of the operations depend on the timing of everything. If you run the CPU a good 6 inches away from it's normal position, you're going to run into HUGE problems.

I dunno if cross talk would be an issue, but I do know that latency would be your biggest issue.....probably a big enough issue not only to cause a performance hit, but probably instability as well. remember, we're not talking millimeters, we're talking several inches. The latency added with the natural resistance of a longer wire might be enough to mess things up aside from performance......

at least....that's my opinion...I'm no computer engineer....yet....

=ACID RAIN=
01-31-04, 07:52 PM
Well thats true, I hadn't thought about latency. Hmmm....

OperativeSix
02-01-04, 02:20 AM
Not to mention the extreme difficulty of taking each pin from the socket to the extension.

Still, if you could pull it off...EXTREMELY cool mod :D .

DrSpanky
02-01-04, 12:11 PM
would be sweet if it worked

tbones1337
02-01-04, 09:13 PM
try it with a old CPU.. it doesnt hurt to try.. unless you shock yourself with 120V

dreadlord79
02-02-04, 01:25 AM
You could always use the existing socket and desolder it from the mobo. It would be best to have a professional reworking solder unit. These are used by circuit board manufactors to rework boards. Others on the board probably know what I am talking about and can give the exact name. After you get it off (if the board isn't dead) you could then run wires from the socket to the mobo holes. Just a thought if you are really serious.

=ACID RAIN=
02-02-04, 04:55 PM
Thanks for your ideas guys :)

I have an old K6-2 machine, so I might try it and see what happens. If it aint broke, tweak it till it is ;)

furymouse
02-04-04, 05:57 AM
Here is something that might help.
If even for just moral supprot.
AGP extender (http://forums.bit-tech.net/showthread.php?t=50470)
Might have diff probs, but hey, they said it couldnt work either.

jacheatamobits
02-04-04, 02:06 PM
heres what he used ^^^ (made)

http://www.scifikg.com/albums/album06/Picture_061.sized.jpg

if you get this to work (which i believe you will) im sure others will follow, and maybe even improve the idea...

much props

brakezone
02-04-04, 09:06 PM
Latency is definitely an issue as electicity travels down 6 inches of wire in around a nanosecond. 1hz is 1 million cycles per second so 3ghz processor ran down 6 inches of wire would cause differences in latency and resistances. a few inches may not seriously effect performance or stability, but seriously with processors this day in age, shorter is better. Lets say its 6 inches the signals might arrive within a specified time and the system will remain okay! What if its to long... the processor is miles away, the signals go to the processor and get there a few nanoseconds later, and the answers start returning after the computer has already decided that it has no processor in it. a 3ghz processor cycles roughly 3 times a nanosecond... Probably best to keep those wires short! weather it would work or not- if done well, would depend upon the timings of the rest of the system and if the additional time was outside the specification of the wait states. i'm just making this all up and guessing.... but geez even if it doesn't work somebody at least give it a try imagine having a giant heatsink on both sides of your chip!

Hexnan
02-05-04, 05:30 PM
solve those latency problems by using fibre optics!

wasnt intel or a sister company of intel working on a fibre optic processor?


Hex