I cant even believe this issue is even up for discussion. The speed of any electrical signal is only as fast as the capacitance of its substrate. And inductance too. NOT by sockets, etc...
This. I mean, really, come on people.
The relationship between the CPU and the rest of a computer isn't like, say, the relationship between an engine and the rest of the car. You're acting as if providing a "better" connection between the CPU pins and the motherboard will be akin to boring out the intake manifold or unkinking the exhaust to eek a few more horsepower out of a given engine block. (Or to use another analogy, like substituting a larger diameter hose between a faucet and a sprinkler to get it to throw a little further.) A 68030 is a state machine synchronized with an external clock input, and it *always* does the same amount of work in response to a single tick of the clock.
*IT'S DIGITAL, FOR CRYING OUT LOUD*. If the connection to the motherboard/socket is good enough for the signal to get through (without excessive noise or interference) it's going to work exactly the same. If the connection
were so bad that the CPU were missing clock beats then the computer's not going to work, because the CPU will fall out of sync with signals generated by other devices on the motherboard, like RAM, and
CRASH. That's it, period. End of story.
If any measurable difference is observed between two SE/30 motherboards, one with a socketed CPU and the other soldered, it's because of some other difference, like slightly variant clock crystals. For people tossing out "trace length" as justification for thinking there's something to this:
A: At the speed electricity travels in a wire you need about a foot of distance to cause a one nanosecond delay. So the extra, what, two millimeters you get with a socket vs. soldered is *incredibly insignificant*.
B: Remember that "clock" thing? Imagine, if you will, a square dance being held in a gigantic hall. What the dancers do while dancing is dictated by what they hear from a caller standing on a stage at one end of the hall, who's shouting into a microphone connected to a single powerful PA speaker. (The hall doesn't have speakers anywhere else, all the sound is coming from the one spot.) In theory, if you were to view the dance from above (and again, assuming the hall is absolutely huge) you might be able to perceive that the dancers at the far end of the hall are dancing *very slightly* out of sync with the ones right in front of the stage. However, and this is the important thing:
they're not dancing any slower. They respond when they hear the call exactly like the ones in the front and do so at the same speed. They just hear it a little later.
When translating this to a computer: Unless the distance from the CPU to the devices it's talking to on the motherboard is so far that the transit time of the signals consumes all the "margin of error" there is for the devices to respond to each other the computer will behave
exactly the same. If the delay does exceed that boundary then the computer won't "run slower", it simply won't work. (Unless the clock speed were reduced to the point that the delay was once again acceptable.)
My theory is because the pins on the socketed one are gold plated...Should make a great electrical contact.
Thought processes like this are why outfits like Monster Cable succeed in milking idiots $50 a shot for "gold-plated gas-sealed made-by-dwarves-in-the-Black-Forest" RCA cables for digital audio. Our bits are cleaner so they sound better!
