Printer-friendly version
Author:
Taxonomy upgrade extras:
Um' well, I understand the idea behind overclocking a CPU and infact used a machine for several years that had been overclocked by a friend of mine who was an INTEL engineer. That was 1998 or so. So here is this video, but I don't understand what they are saying about it. Does it mean that they are running a 300 Hz chip at 5000 Hz?
http://www.custompcblog.com/industry-news/what-is-coltan-tan...
Much Peace
Gwen
Comments
Re: I don't understand this?
Yes, this is exactly what they are saying. The faster a chip runs the more heat it produces. The limiting factor in overclocking is the temperature of the chip, go too fast and you will burn it out. By using liquid nitrogen to cool the chip you can get massive speed increases while keeping the core temperature within exceptable limits. While a nice demo of what can be achieved it isn't exactly a practical exercise, how many people have a ready supply of liquid nitrogen?
Strange you should ask.
I myself am going to be an intern at the lab that actually utilises liquid hydrogen on everyday basis and even liquid helium on a... biweekly basis. That's what you get for NMR spectroscopy.
Faraway
Big Closet Top Shelf
Where you can fool around like you want to and most you get is some bemused good ribbing!
Faraway
Big Closet Top Shelf
Where you can fool around like you want to and most you get is some bemused good ribbing!
Liquid Nitrogen.
When I was working in Industry, we used Liquid Nitrogen as part of a test heat treating process. It was quite fun actually. You could dribble silicon cement into it and make hard little balls, and when we threw them, they could go off like firecrackers. :)
Gwen
You folks are making me want
You folks are making me want to get back into studying science again, calculus be damned ;-) (HATE MATH! ... But I'm wondering if I could muddle through now >_>)
Not quite
The CPU wasn't originally a 300 MHz one. It's a Pentium 4, the slowest of which is 1.4 GHz. They probably took one of the late-model P4s (the ones with over 3 GHz clock, and after Intel managed to partially solve the notorious P4 overheating problem) and raised it to 5 GHz-plus.
The 300 MHz referred to the front-side bus, that is, the data channel the CPU uses to communicate with the rest of the computer. Since P4s are multiplier-locked, you have to increase the FSB rate to increase the CPU clock. Which means that the rest of the board will heat up quite a bit too, so you have to increase the cooling for the chipset.
It's not practical, it's a proof-of-concept thing and a bragging rights thing. But it's impressive nonetheless.
Temperature not the only limiting factor
When over-clocking a chip or even designing a chip for digital operation (from cutoff to saturation on the transistors) in the GHz (giga hertz (10^9 hertz)) region one needs to not only consider the thermal dissipation (which can be significant) but also the switching speed of the actual transistors (from cutoff to saturation) great advances have been made in this area but there is a theoretical upper limit to the performance of a given chip due to this factor.
Or at least that is what my digital systems course at uni lead me to believe...
Knofster
You are correct
If it takes longer than a quarter of a cycle or so to go from saturation to cutoff, you're simply not going to get the signal through.
As far as heat goes... When a transistor is at cutoff, it is dissipating very little power because there is practically no current flowing. When it is at saturation, it is dissipating somewhat more power, but still very little if it's designed well. There is little resistance, so most of the power is consumed by the load.
In the case of digital circuitry, the load is the gate of the next transistor, so there is only a very short pulse of current -- enough to fill the stray capacitance of the load (the gate of the next transistor.)
When the transistor is in the process of switching, it goes from nearly infinite resistance (open circuit) to very low resistance. During that transition, power is consumed and heat is generated.
In essence, each time the gate is switched, a certain fixed amount of heat is generated. Therefore, the more times you switch it per second, the more heat is generated per second.
By the way, a transistor that switches faster will generate less heat per switching. As the chips improve, it's possible to switch them faster.
The next generation of semiconductors will probably be nanotubes. The next generation of logic may be optical, or it may be superconducting switches -- perhaps josephson junctions.