Advanced mod- Socket extension?

Aristarchus

Limp Gawd
Joined
Mar 24, 2001
Messages
263
How exactly are we defining "work" here? If you're talking about dropping the clock speed to something low or sub- MHz, then yes, perhaps it would. But in so doing, you'd open up a whole new can of worms with your memory refresh rates, and whatnot all. And all in all, you'd end up with something that's more or less a glorified PIC, and many hours of your life gone.

The pins on the processor are not simple circuits like the sort of thing that runs your toaster. Even at antiquated clock speeds, those pins and traces behave as transmission lines, and you enter an entirely new realm of evilness, because the capacitive coupling between lines and ground will completely jack up the signalling process, due to latencies and crosstalk, yes, but also do to varied termination impedances. It's likely that what you're discussing will give signal paths having step responses that look like a bell at any substantial clock speed, so the device at the receiving end of the line won't have a clue as to wether it's seeing a 1 or a 0. If you want to try it, I'd use something built before PCI slots were invented. Also, please do understand that this will take up a hell of a lot of time, and most likely give a mediocre result at best. So if you want to do it, more power to ya, and don't grab the iron by the wrong end.

Oh, and have fun, eh.
 

plot

Supreme [H]ardness
Joined
Apr 3, 2002
Messages
4,794
there won't be any latency issues because there wont be any signals period... MHz and GHz flying through wires create an extremely large amount of noise. just the processor itself sitting in it's socket is a source for alot of noise, with 9inch antanneas sticking out of each pin... holy shit. not gonna work
 

gee

Electronics Wizard
Joined
Aug 13, 2000
Messages
2,119
Having designed PCBs that involve high speed FPGA/DSP parts, near-GHz clock rates, and so forth... I think I can confdently say - it's not possible.

Here's why.

1. power problems - adding those wires will add lots of inductance and resistance. You'll need to add a ton of decoupling capacitors at the processor itself just to make it stable... as well, you'll need to add termination resistors and other things at the pins to keep overshoot/undershoot on data signals from blowing stuff up.

2. latency - electricity travels awfully fast, but not fast enough - adding several inches of extra distance between a north bridge and CPU, or between a CPU and its memory, etc... can introduce a time delay high enough to cause all sorts of bad things to happen.

Something like a GTL+ P3 uses a packet based bus to get around a lot of the timing problems, but things like clock skew will still bite you in the ass. A regular "local bus" CPU (eg anything socket 7) just won't work.

3. That's a lot of fucking wires.

Feel free to do it with something like a 286 though... I'd draw the line at a 386.
 

Jaden

Weaksauce
Joined
Jul 2, 2003
Messages
117
Here is my two cents:

There will be noise between the wires. There is noise in literally every wire in your computer. EE's working on motherboard circuits design the wires differently. Instead of "just resistance," it's modeled as an RLC circuit (inductances because the wires turn at times, and there is capacitance when there are wires with an insulator between them - the layers of the motherboard = insulators). A little noise, however, is not truly a big deal with you convert it to 0's and 1's.

There are other problems though, mainly the soldering. I cannot imagine soldering all of the pins on a cpu. The probability of two of them touching is extremely high; with that said, while I would like to see this mod, it's probably not worth doing. Not to mention... where exactly would you mount the cpu? Regardless of where you put it, a heatsink/fan will have to be on top. I think doing this to a video card would be much more rewarding :)

Basically I don't believe it's worth the time arguing over this. It can be done, but it has a very bad probability of working well/at all.
 

hctimo787

n00b
Joined
Aug 12, 2004
Messages
41
I only read half the thread, I just wanted to say, at the very least, it would be excellent soldering practice. Wuddya got to lose?
 

mattg2k4

Gawd
Joined
Sep 24, 2002
Messages
829
He'll lose a couple of hours, a bit of solder, and the tip on his iron will get that much closer to replacement time.

It won't work, so soldering practice really is all that wll come of it.
 
Top