Advanced mod- Socket extension?

Vegeta

Gawd
Joined
May 12, 2002
Messages
595
I was wondering if anyone would think it possible to take an older (aka cheap) motehrboard and processor (K6 266 Socket 7 era) and mod the socket to extend it via small bundled wires in order to place it somewhere else in the case. This would involve soldering a wire to every pin on the processor and then locking the other end of that wire into the socket. Doing this could make for some cool mods. I was wondering if it would cause problems with either voltage drop and/or interference in between the wires. They would all be the same length/grade wire, probably bought precut from RadioShack like they sell for kits. I'd use the smae guage wire used in IDE cables for rough comparison. THey wouldn't need to be any longer than say, 9 inches.

Is this possible? I think it would be an extremely amazing mod.

If not hte processor, then how about an PCI or ISA slot... remove the slot and solder wires to move the slot some other postion in the case.
 
not a very feasable mod IMO, since you'll run into some outrageous timing and latency issues doing that....not to mention the shear number of wires to deal with...

it'd be worse than a 20year old telco cross-connect box.
 
speedracer089 said:
even if you could do this attaching a heatsink or anything to cool it would be a hassel
i really don't think that's a concern of Vegeta's.

I personally would love to see that done. It reminds me of the movie pi.

I've also always wondered if that's possible. But i dont' have the soldering ability to do that.

if you do attempt this, be sure and post pics and what not :)
 
Possible with a lot of work.

Effect: Fucks up CPU speed from cpu to mobo.....you'll be further away and I got $5 that says it'll screw with the signals going through the wires. Not only would it be a bad idea, what "cool" mods could you do? mount it externally? Uhm right.


I could see extending AGP and PCI cards to mount them differently but they already sell stuff like that (even though its a rare thing to come across).
 
Qtip42 said:
Possible with a lot of work.

Effect: Fucks up CPU speed from cpu to mobo.....you'll be further away and I got $5 that says it'll screw with the signals going through the wires. Not only would it be a bad idea, what "cool" mods could you do? mount it externally? Uhm right.
what ever happened to people modding computers in outrageous ways!? I think the concept alone is great. Even though it's a very improbable mod, and would most likely fail, i think it's a great idea!

I could see extending AGP and PCI cards to mount them differently but they already sell stuff like that (even though its a rare thing to come across).
they're not rare to come across, just no one ever uses them in mods, so we don't really see them that often :( Buying them definitely isn't a difficult task however.
 
NickTheNut said:
what ever happened to people modding computers in outrageous ways!? I think the concept alone is great. Even though it's a very improbably mod, and would most likely fail, i think it's a great idea!


they're not rare to come across, just no one ever uses them in mods, so we don't really see them that often :( Buying them definitely isn't a difficult task however.


I'm not knockin doing it......just dont see the point if it wont work. And i said they're rare because you dont find them in your everyday modding stores.
 
that won't work ( read : " lengthening the signal lines = the crap " )
All you might achieve with such a mod would be a monstruous radio wave field generator inside your case ( not really powerful, but nor really insterresting, 'cept for a science project XD

edit : did I mention the risk of starting a fire ? most recent CPUs draw around 50-60 Amps for power ... I don't thing you wanna go for space tech wiring, do you ? :D
 
Qtip42 said:
Not only would it be a bad idea, what "cool" mods could you do? mount it externally? Uhm right.

Dude wth are you on. Use your imagination. Imagine 259 wires going into the "central core" or something. With right lighting, and clear case you can make it a crazy ass mod indeed.

Imo there are 2 ways to pull this off. One is to fake it. Use a slim waterblock on the real cpu, somehow mask it, attach a socket onto the waterblock, attach your crazy extention cable. and mount a fake cpu in the middle. Just think about. Clear/tinted acrylic cube case. CPU (fake) mounted in the middle. Supported by 8 wires, coming from each corner of the cube to the cpu. For wiring you could use a crazy telco bundle like you have seen in those switching boxes. Or, use fiber type wiring, but like the type that they use in those lamps, that each fiber glows. Those could be made to look very scifi'ish.

Second way to do this, is do the mod for real. But you would have to drop the mhz to a crawl. Literally. Probably ~ 10mhz is all you will pull off. It is duable all right. How do you think chips are tested out before they are taped out? The entire cpu core is "wired" up. It works 100% but at like 2-3 mhz.
 
I dont see the Real thing as pheasable but on the subject of hiding it you chould make a fake MB tray to attach over the real one and then mod the fake one up like you want to ...


t10 said:
Dude wth are you on. Use your imagination. Imagine 259 wires going into the "central core" or something. With right lighting, and clear case you can make it a crazy ass mod indeed.

Imo there are 2 ways to pull this off. One is to fake it. Use a slim waterblock on the real cpu, somehow mask it, attach a socket onto the waterblock, attach your crazy extention cable. and mount a fake cpu in the middle. Just think about. Clear/tinted acrylic cube case. CPU (fake) mounted in the middle. Supported by 8 wires, coming from each corner of the cube to the cpu. For wiring you could use a crazy telco bundle like you have seen in those switching boxes. Or, use fiber type wiring, but like the type that they use in those lamps, that each fiber glows. Those could be made to look very scifi'ish.

Second way to do this, is do the mod for real. But you would have to drop the mhz to a crawl. Literally. Probably ~ 10mhz is all you will pull off. It is duable all right. How do you think chips are tested out before they are taped out? The entire cpu core is "wired" up. It works 100% but at like 2-3 mhz.
 
Well a friend of mine just happened to pickup a grab bag of old processor today.. I think some K6's and a few celerons and maybe an old pentium 1... plus I know we have a Pentium 1 laying aorudn here somewhere... Now there are only like 186 pins on a P1 I belive... It hs been a while since I have been around one... Both myself and my friend are experienced in soldering and have soldered such things as IC pins before... we could split the time taken up between us....

As for the latency issue....

look at how far away the processor is from the northbridge on some motehrboards of that era... about 3 inches or so... the current in the traces has to travel that far.. Plus the old Pentium 1's use either 3.3v or 5v belive it or not... so signal strength is nothing.. plus those chips aren't built with frontsides as fast as todays... depending on how patient the northbridge on the motehrboard is.. it should work... just possible a little unstable or slow. The only thing that bothers me is all those wires so close to each other. I could see the processor possibly dropping cycles depending on how well insulated the wires are.

As for modding possibilities... those of you who said it would be pointless... where is your imagination? Who wouldn't think it would be cool to look in a case and see a large bundle of wires leading to the processor? We were thinking put a window int eh front of an old case and actually stick the heatsink through a hole in the window..with all the wires leadign back a few inches to the socket...

Or, if it runs good.. try going without a HSF at all... I know some of you would advise not to do this but I used to own a DX4 100mhz laptop proc that had no heatsink after it had fallen off... it ran in my laptop perfectly without any hangs. My friend ran his K6 380 that was overclocked to 416 on passive cooling for 4 hours while downloading windows updates...

This is why me and my friend are so interested in older processors.. they are extremely durable compared to modern processors. We are planning to take a 266 K6 and pop the heatspreader of and take a miniature heatsink scrapped off an old IBM os/2 machine that seems like it would fit perfectly on JUST the core and surrounding area.. leaving the rest of the chip exposed... and see if it will last... now how cool would it be to see THAT in a window?

Sometimes ya gotta go old school...
 
why not make an extension?... like, lets say you're extending for P1 chips... you make a replica of the pins and with the wires, hook them up to another desoldered ZIF socket of a P1 so that the copied P1 is the plug into the real ZIF socket and then the extra one is for the real processor that you lock on and for the heatsink to clip on
 
Okay. You obviously aren't anything close to an electrical engineer. Neither am I, but I do know that chip makers and motherboard makers pay a damn lot of people a damn lot of money to engineer these things. Every last detail is examined and engineered. Latency is a HUGE issue. Motherboards are designed to work with the exact traces that are on the motherboard. It isn't a matter of overall distance so much as it is a matter of changing the distance. Even older processors rely heavily on trace length. I don't even want to get in to the issue of crosstalk. Basically what I am saying is there is no way this is going to happen with any modern processor. I'd be willing to bet you could pull it off with a z80 or an 8088 but I'd say pretty much anything post-pentium era isn't going to work worth shit.

btw, insulation isn't going to mean jack shit with the wires other than if they are touching or not. The big thing is shielding, which seems quite impractical for a few hundred thin wires.
 
I'd have to say it's not going to work most likely.
I'd have to say faking it is better and less work probably.

heh I'm a computer engineering student.. but I'm just started taking all my circuits classes and stuff .. so I really can't comment with any expertise. However I think this normally would definately not work.. but with these old chips who knows? Wouldn't it be better though if you could fake it so well that no one could tell?
 
really wont work. some will remember the apex dvd hacks a while back where similar things were done. even in that there were problems mounting different things less than 1/4" higher than where they were supposed to be.
 
OK so it may not work... but there is the possibility that it will... I do understand the motherboard manufactuers engineer everything to be perfect... but that is so the chip runs perfectly.. we dont need this thing to pass Prime95 or anything... just boot up. even if it is wildly unstable and slow as hell, as long as it boots and recognizes the processor I will consider it a success. This is another reason We will be using a old Pentium 1... older technology isn't as picky when it comes to things like this. I ran a 486 chip that had a pin missing for a few hours before it locked up and gave me a beep of death.

As for crosstalk is concerned... IDE cables don't suffer from crosstalk... allthough I am now recalling that there is a ground wire in between each data wire on IDE drives for that reason...

Oh well. I will try it and let you guys know. Any other ideas for something like this? maybe take some old D-RAM and solder wires to in and put ram in a window? Now that would be a neat looking mod. Again with the latency issues... well we will see.
 
... I do know that chip makers and motherboard makers pay a damn lot of people a damn lot of money to engineer these things. Every last detail is examined and engineered. Latency is a HUGE issue.

...

This guy knows what he's talking about. Despite the inflammatory tone, he speaks the truth. One issue he did not mention is power. Even old pentiums take a fair number of amps, which all have to go through those pins into the CPU. I don't want to be a wet blanket, but jpmkm is right.

Dan has a good (albeit short) response to someone with the same question here.

I personally kinda like the idea of faking it. Say, take a cooler with a long vertical heat pipe, surround it with wires (to hide the real CPU on the MB), and on the other end put the old CPU with a big hole chopped in the middle (to let the heatpipe through) and a heatsink on top of the heatpipe. The wires make it look like an extension cord for the CPU, and you cool your working CPU with the heatsink on the non-working CPU.

Sorry if that didn't make any sense. It's bedtime here.
 
what if you were to create a type of optoisolator for each essential pin (Address bus, data bus, sideband signals. you get a crate of LEDS and several meters of fiber optic cable.you then mount the processor on a seprate board (multi-layer of course) and make an array of these LEDs. you then use pull up resistors to bring the current output from the the processor address, data, and sideband signals to ~3v that way you can power the LEDs. you then have a few (300 or so) phototransistors. and when everything is wired up you can run the processor at some great distance (as long as the fiber optic cable will let you). but you will end up with alot of fiber optics and a massive array of SMT LEDs(or lasers), it would be cool though to have a few of these fibers unsleved so you can see the light(at least some of it) anyway back to topic.... If you use light there will be no cross talk, latency or any of the like ( as long as your leds and phototransistors can keep up with the processor )
 
x1600c said:
what if you were to create a type of optoisolator for each essential pin (Address bus, data bus, sideband signals. you get a crate of LEDS and several meters of fiber optic cable.you then mount the processor on a seprate board (multi-layer of course) and make an array of these LEDs. you then use pull up resistors to bring the current output from the the processor address, data, and sideband signals to ~3v that way you can power the LEDs. you then have a few (300 or so) phototransistors. and when everything is wired up you can run the processor at some great distance (as long as the fiber optic cable will let you). but you will end up with alot of fiber optics and a massive array of SMT LEDs(or lasers), it would be cool though to have a few of these fibers unsleved so you can see the light(at least some of it) anyway back to topic.... If you use light there will be no cross talk, latency or any of the like ( as long as your leds and phototransistors can keep up with the processor )
How do you figure there will be no latency?
 
speed of light rather than electrons.. however.. I'm not so sure those leds light up instantly?

edit: oh yeah how much space will like 300 LEDs take up? and what are you going to do after all your signals are turned into light?

Not to mention there's more problems probably with this idea then wiring it straight
 
The speed advantage of light over electricity(if there is any noticable advantage) will be completely negated by the conversion of electricity to light to electricity. Light still takes time to travel so there will still be a latency issue. Sure it looks fast for us, but you must consider that the light will be going on and off a few hundred million times a second. Also, one probably won't need nearly that many LEDs. A lot of the pins on a processor are power, which cannot be transfered by light.

This LED idea could work, but it would probably involve designing a new motherboard and processor to do it.
 
Tos said:
did I mention the risk of starting a fire ? most recent CPUs draw around 50-60 Amps for power ... I don't thing you wanna go for space tech wiring, do you ? :D

most of the P1 era chips used less than 10 amps at either 3.3 or 2.9v

and i doubt wiring for that low of a voltage (albiet high amperage) would be that difficult...

after all the traces on the mobo arnt that big either
 
Qtip42 said:
ouch now thats a spendy price for an extension

well its the first one i could find in google, anyway thats not the point and you know it :p
 
Hey, if you happen to accomplish it, or even an attempt at it, post some pics, okay? :)

I just looked at an old P1 chip I have here, and it has something like ~332 pins, give or take, if my math is correct. After soldering all of those, I wanna see how bad your camera hand shakes!! LOL :p
 
yeah but remeber this: besides extra current draw, PII runs at about 150-300mhz. our prossesers run at 2.5-3 GHZ. signals running thrugh long wires at that speed are going to be very vunrable to noise. and timming issues, the CPU, north/south bridge and ram shift data in and out at up to 800 mhz ( fsb ). the timeings for all of these things has to be exact, otherwise bad things happen. this is why overclocking sometimes messes things up.

so its a great idea, if you had a 486 i would say go for it. or even if you had a dead prosseser. but with your live CPU i would deff say nogo
 
Hate to be a wet blanket, but even if the latency wouldnt be a problem , would the fact that
a man with a soldering iron can never solder in such a way that all the cables are exactly the same length be of any use/consideration

Even if all the cables are cut/bought at exactly the same length, there is no way someone except for that dude on the homeshopping network, who fixes a burned out car with some Astonish and a tin of Motor-up extreme and McGuyver/B.A. can solder all those wires to those pins at the same spot.

Even if U solder them all on the same spots on the pins then the soldering can/will result in resistance differances (rosin core solder isnt always the same throughout the roll, i.e. there can be 99 percent solder in it, but a cm further it can be 98,9 percent solder.)

The cables can also be of differnet quality resulting in different resistance
 
aL Mac said:
resistance wouldn't be an issue

Why not???

I thought that the resistance would increase the time for a signal to get from the socket to the cpu/vice versa thus screwing up the timing of those signals.

i.e. the cpu computer 3 bits of data (a 0,a 1, and a 0 again)
It starts sending that to the mobo/socket, but because of the resistance/different wire lenghts
it doesnt come through in the original 010 but as 001 or 100, isnt that data corruption.
As there are a minimum of a million signals going to/from a proc at any given time, all travelling at the speed of electrons, i think resistance is a major problem.
 
lol good luck doing this mod. :D

Most likely you'll end up with a pile of wire and something that's looks like it could work in bizarro land :)
 
BioHazard.NL said:
Why not???

I thought that the resistance would increase the time for a signal to get from the socket to the cpu/vice versa thus screwing up the timing of those signals.

i.e. the cpu computer 3 bits of data (a 0,a 1, and a 0 again)
It starts sending that to the mobo/socket, but because of the resistance/different wire lenghts
it doesnt come through in the original 010 but as 001 or 100, isnt that data corruption.
As there are a minimum of a million signals going to/from a proc at any given time, all travelling at the speed of electrons, i think resistance is a major problem.
Resistance shouldn't have any effect on latency. Do you have any sources to back up your claim? Wire length will certainly affect transmission times, but resistance differences due to wire length differences will not.

Also, very few CPU data busses transmit data serially. Therefore, while the data may not arrive at the same time(due to wire length differences), the data would not be out of order like you say it would be. It will still cause problems, though. There aren't a million signals at any give time. This would require a million signal lines, which processors don't have. Rather, there are a few(relatively) signals, but those are signaled millions of times a second. And electrons actually travel surprisingly slowly through wires. The signal still moves quite fast, but the electrons themselves move slowly. Again, though, resistance should not really affect transmission speed. It will simply affect how much voltage is required to drive a certain current through the wire(I hope I got that right).
 
BioHazard.NL said:
I thought that the resistance would increase the time for a signal to get from the socket to the cpu/vice versa thus screwing up the timing of those signals.
It doesn't. Trust me, resistance has nothing to do with timing. It WILL result in signal degradation, but your signals will be altered by capacitance (the resistance to voltage changes) and inductance (the resistance to current changes). Look up LC and RLC circuits in any basic college level physics book.

-special [k]
 
Cool idea, lots of work though.
To actually answer your questions, instead of saying "ooo thats a bad idea, ooo that wont work, ooo why would you do that, yadda yadda". This is the [H] people, keep that negativity to yourself. I for one applaud someone for doing something silly like this, just for the hell of it.

1. EMI - electromagentic interference from the CPU is real. Unshielded wires could cause problems, either by generating or picking up EMI. This is easily solved with some shielding. The standard rubber/plastic around the wires won't cut it.

2. Interference(cross talk) between the wires - as long as they are insulated, there wont be interference

3. Voltage drop - depending on the length of wires, this could be a problem. Make sure you use good quality wires. Unfortunately, theres not much of a way to predict what this will do...you could use an Ohm meter and check the resistance of your wires, get it as low as possible.

Last thing...soldering all those wires will be a nightmare. Get a good quality soldering iron, good solder, and a nice clean well-lit work area. Some sort of magnifiying glass thing will help too. Don't mess the pins up either...ROFL
 
Steel Chicken said:
Cool idea, lots of work though.
To actually answer your questions, instead of saying "ooo thats a bad idea, ooo that wont work, ooo why would you do that, yadda yadda". This is the [H] people, keep that negativity to yourself. I for one applaud someone for doing something silly like this, just for the hell of it.
Negativity? Explaining why something won't work is negativity? The OP asked if it would work and we answered the question as best we could by explaining why it wouldn't work.

1. EMI - electromagentic interference from the CPU is real. Unshielded wires could cause problems, either by generating or picking up EMI. This is easily solved with some shielding. The standard rubber/plastic around the wires won't cut it.
Very good point, but it complicates things greatly(sheilded wires are quite a bit thicker and there will be 1.5 times as many connections to solder).

2. Interference(cross talk) between the wires - as long as they are insulated, there wont be interference
Now that just contradicts what you say in 1.

3. Voltage drop - depending on the length of wires, this could be a problem. Make sure you use good quality wires. Unfortunately, theres not much of a way to predict what this will do...you could use an Ohm meter and check the resistance of your wires, get it as low as possible.
This is especially important considering how much power a processor draws.

I still feel the biggest problem will be latency, though.
 
jpmkm said:
Negativity? Explaining why something won't work is negativity? The OP asked if it would work and we answered the question as best we could by explaining why it wouldn't work.
[/b]
It will work, IMO but theres no way of knowing for sure until someone tries it. Now get your trapper keeper, because its times for school, and you obviously have no idea what you are talking about.

Very good point, but it complicates things greatly(sheilded wires are quite a bit thicker and there will be 1.5 times as many connections to solder).
you dont need to shield the individual wires (EMI wise, not current wise) from each other. You only need one shilded bundles around the whole mass of wires. EMI <> current shorting out. Are the individual pins on the cpu shielded from each other, EMI-wise? No. Why would need to shield the wires from each other?

Now that just contradicts what you say in 1.
No because, EMI <> current shorting out . The individual wires need insulation to make sure they dont physical come in contact with each other (bare wire), but as far as EMI shielding goes, one large bundle around all of the wires will handle it. You have no clue what you are talking about.

This is especially important considering how much power a processor draws.
the wires would have to be roughly the same diameter and therefore power capacity as those tiny little pins on the bottom of the chip. Not exactly 10 gauge are they? You ever used a pelt? You know much power they draw? How small their wires are? Wire size wont be a big deal as long as they are not too small. A careful person would identify which CPU pins draw power and are ground, figure out how much power and then get the minimum gauge wire. The rest of the data signal pins could be very small.

I still feel the biggest problem will be latency, though.
latency as compared to what? If you took all of the traces inside the cpu and on the motherboard and layed them out in a line, they would not be that much smaller than the roughly nine inches compared to the wires. Also, latency is not much of a big deal when your talking about electricity, whose impulse moves at the speed of light. And even if it does adds a little latency, its besides the point, isn't?

Were just trying to see if ti will WORK not if it will work perfectly like you never did it.

Im so sick of morons who don't know anything (such as the difference between electrical shorts and EMI) trying to tell people something can't be done. If you dont know what your talking about, stop giving your made-up advice to people who are more adventurous then you are.
 
You never said a goddamn thing about shorts. You said that insulation will protect against interference, which is bullshit. The possibility of crosstalk among wires is very real. ATA66 cables have a ground wire between each signal wire to guard against crosstalk, and IDE transfers are considerably slower than some of the busses from the CPU to other devices on the motherboard.

Now for the power requirements. You said that voltage drop was important to consider and I supported your argument. Why are you arguing with me?

Now the latency issue. REGARDLESS of how long the traces on the motherboard are, changing the length of those paths changes the characteristics of how signals travel through the path. I don't give a shit if the motherboard traces are five inches or five feet; the issue is that everything on the motherboard was precisely designed for that specific length of trace and changing those lengths could have undesired effects.
 
I did say something about shorts, but I after re-reading my post I can see why you would think I didn't. The fault in communication was mine.

I am arguing with you because you said this couldn't be done.
My point is i think it *can* be done if you understand whats involved and plan accordingly.

Your point about latencies, I understand what your saying, I simply disagree with you. If you slow down you RAM timings, does the computer suddenly stop working because of "latency" issues? If you turn off your CPU's cache via BIOS, which makes data travelling back forth to RAM and the north/south bridges much slower, magicaly cause the computer to not work because of latency issues? NO If you were to put in a slower CPU (supported by the mobo) would the computer suddendly stop working? NO

Is latency real? YES, will adding these wires increase latency for some operations? YES
will the computer suddendly not work anymore because of it? I BELEIVE NO. I beleive this because many of the components in the computer are designed to handle various latencies for various transmissions of information...why is this any different?!? Anytime you swap a CPU, or over or underclock it, or put devices on the PCI bus, or change the speed./latencies of your ram, you are changing system Latencies, which the computer is designed to dynamically handle. Thats my point.

But I guess he should not try, because you dont think it can be done.
 
this will never work with a modern prossesor.

if you want the effect, get a dead moden prossesor, and then use a Basic Stamp or something to like up leds at random... haveing a actual prossesor running off fiber or anything else is just pointless because having 300+ leds on one board (even SMT LEDs are much much bigger then prossesor pins) would never fit, you would need your own fiber (it just doesnt majicly sit there, you have to mount it all) and then the light->digital isnt that simple, you cant just solder up 300+ photo transistors, you need suffecent power for each and ground for each and agein, you would have to mount them all somehow and secure the fiber...

and if you cut open one or two fibers (Assuming that you soldered perfectly and mounted all the fiber perfectly) then you would just be messing up the prosseser, and it would be "blinking" too fast to even see... ie it would just look dim/on.

so in summery, dont do it its stupid. i am not oposed to working art out of a dead prosseser, but a functional prosseser = bad idea
 
I think you are a bit confused as to what exactly constitutes latency and how a computer actually works. Your comment about turning off a cpu cache slowing down data transfer between memory and the memory controller really shows that you have absolutely no idea how a computer works. A CPU cache DOES NOT affect transfer speed of ANYTHING WHATSOEVER. The only reason a cpu cache speeds up anything is because the cpu can grab things out of the cache(which is really really fast) instead of having to go all the way to the memory controller to ask for it(which is slower). Also, changing the speed of ram does not change how long the signals take to travel through the wires. Changing the speed of anything does not change the time it takes for signals to travel through wires. The only thing anyone can do to change the time it takes for signals to travel through wires is to modify the wires by, i don't know, making them longer.

Motherboards and components are designed to work with a certain range of latencies because there are always slight differences in manufacturing, but we're talking about a range of maybe a few millimeters difference in length.

So to sum it up, it's easy for you to say I don't know what I am talking about when you obviously have no idea what you are talking about.
 
Back
Top