Slashdot Mirror
More Super Cool Overclocking
octools.com has a followup to a story we linked there a couple of months ago where they submerged a motherboard in nitrogen cooled flourinet, and overclocked the hell out of the chips. Well, they're back with an extensive photo documentary of
the sequel where they try to take it below zero, and clock things over a gigahertz. You probably shouldn't be trying this at home, but it sure is fun to see on a Web page.
The fact remains that the cpu's that most overclockers use are very cheap in comparison with the latest high end chips. For a few extra dollars spent on a good motherboard and a Golden Orb cooler overclockers are safely pushing 600 MHz Celerons to 850 Mhz and beyond. With this sort of price for performance, one might ask, "Why wouldn't you overclock?"
That said, this article was obviously done just for the sheer geek of it and power to them. The advances in cooling could easily be used in future production machines.
If you have to ask "Why overclock?" then you are obviously not the type who takes every new toy in your home apart as soon as it comes in the door just to see how it works and how you can screw with it.
IANAOC. Seems they had success when everything was uniformly cooled with the Flourinert and dry ice. Problems arose when they used the Liquid Nitrogen on JUST the processor. So, the CPU could go faster than the support chips which were not similarly cooled? Maybe the video card, though now running with the same bus speed as the LN-cooled CPU could not operate at those speeds without also being cooled to LN temps. They mentioned:
CPU could handle it okay, but the rest of the system was breaking down. Gotta cool the WHOLE thing, or else your system is only as fast as your slowest link.
So, for MISSION: SUBMERSIBLE 3, I'd like to see them try immersing the ENTIRE RIG in LN, with good-sized heat sinks on the CPU, video card... everywhere and THEN see how it worked. The major concern I'd have would be they migh be encountering a race condition between components that would never arise at conventional speeds.
Offtopic, but here's an idea of what they could have done with the LN when they were done with THAT experiment. (I attended a party in college where we actually DID this. IIRC, It was some time around 1979 or 1980.) Use the excess LN to freeze some vodka in ice cube trays! The vodka will easily freeze at those temps... voila! VodCubes! Take a couple VodCubes, drop 'em in a cup of collins mixer, wait for the VodCubes to stop dancing around on the comparitively hot collins mixer, and enjoy your vodka collins! Looking back, I wish we had tried it with orange juice -- could have called it a frozen screwdriver!
Recently, in one of my EE courses, overclocking was mentioned. It was also mentioned why certain designs can only be pushed so far, and after that they cannot be clocked any higher, regardless of how well they are cooled.
The problem is with propagation delays. Basically, when one part of a digital logic circuit goes high or low, this change takes time to propagate to the rest of the circuit. This is a very small delay, but when you have a circuit as complicated as, say a Pentium III, it can become important. There are small gaps in between parts of the circuit being in different states, and as the clock rate is pushed higher and higher, these gaps become smaller and smaller. If the clock rate is pushed too high, different states of the circuit will overlap, essentially causing it to malfunction. It should be noted that a circuit or portion thereof does not change states (0 to 1 or 1 to 0) instantaneously; this is what allows the overlap that ends overclockability. Propagation delays also explain why, for example, you can't push a C64 to 200 MHz. The circuits in the CPU weren't designed for speeds like that.
=================================
I pledge allegiance to the flag...
of the Corporate States of America...