Slashdot Mirror
PC's Waste Heat Could Add To Processing Power
Urchin writes to tell us that physicists working in a new field called "phononics" claim that waste heat from a processor could actually be used to add to its power. "Crunching data coded using photons — photonic computing — is one example, and in 2007 researchers built the first workable optical transistor. But now the idea of computing using heat flow is gaining popularity among applied physicists. Heat travels through solid materials by means of phonons — ripples of vibration passing through a series of atoms. Those ripples can be used to send and store data in digital form: one temperature is read as 0 or 'off' while a second, higher temperature is interpreted as 1 or 'on.' Provided that the thermal memory is well insulated, it can keep its temperature — and data — intact for a long time."
If I understand this properly (and it's not 100% guaranteed that I do), this sounds like an excessively complicated solution that would yield relatively little benefit. The "sandwich" idea from TFA sounds especially counterproductive, if external power is required to keep the hot side hit & the cold side cold.
Instead of trying to harness waste heat to eke out a fraction of a percent of extra processing power, here's an idea: how about sucking that waste heat into a small insulated pipe with a low-voltage van, and running that pipe down to my feet? It's very cold near the floor of my apartment, and some warm air aimed at my tootsies would be greatly appreciated while I use my computer.
Maybe this pipe could have a little door I could close in the summer, when the additional warmth would be less welcome.
Just once I'd like someone to call me 'Sir' without adding 'You're making a scene.'
Seems like a solid state thermocouple might be easier to use. I'd like to see some sort of heat pipe from the case to one, then use that output to power the screen (maybe not the main one but a smaller backup little screen??). I have no idea of the state of the art there though or what sort of useful electricity you might get from one. I have seen a kerosene lantern from Russia that uses a thermocouple to scavenge waste heat from the kero burning to provide power for a table radio.
At one of our compute farms, we actually pipe the waste heat into a local town as a low-cost house-heating solution (think steam-pipes, but lower quality energy.) It works there because even 100 degree hot air is nice to have when the outside temp is 0f.
VTEC stands for Variable Valve Timing and Electronic Lift Control. It allows the engine to adjust the valve timing in real time to provide additional power at higher RPMs. It essentially allows low displacement engines to expand their powerband at very high RPMs.
It does absolutely nothing for ride quality and can be employed at any road speed, provided the transmission allows the engine to operate within the RPM required for VTEC to operate.
If sharing a song makes you a pirate, what do I have to share to be a ninja?
A computation is a process in which we take a memory area that may be in any randomly-chosen state, and reconfigure it to be in one specific state, corresponding to the return value of our computation.
This is a local reduction in entropy - reconfiguring that memory area into a single state out of the many possible. That means work has to be done, and there has to be an increase in the entropy of the Universe at large that at least exceeds the decrease in entropy in the memory chip. And that means heat.
Real Daleks don't climb stairs - they level the building.
At the lowest possible level, we're talking about electrons moving around. Every time you do an instruction, a bunch of electrons have to move from one place to another. On the way, they inevitably bump into things. Whenever that happens, you lose a bit of energy as heat. That's what the oh-so-common equation, P=(I^2)R means. The I is the current (moving electrons), the R is the resistance (things electrons bump into), and the P is the power (energy per second that you lose as heat).
As for what you read about (ir)reversible math, that does contribute as well. Consider a single bit of a register, where a zero is represented as 0V, and a one is represented as 1V. If that bit contains a 0, and I want to make it a 1, I need to pull out a bunch of electrons. If I am going to move the 0 that used to be there someplace else, then I could, theoretically, move the electrons there. In that case, the only loss would come from the very short distance that those electrons move. But if I am just getting rid of the 0, those electrons go to the 1V supply, and get sent back to the power supply and all the way back to the power company*. That's a longer distance, so there's going to be more loss. But even if all the math was reversible (i.e. no bits were ever overwritten, only shuffled around) there would still be loss. That's a purely theoretical system anyway - no one actually shuffles their bits around like that.
Hopefully that answers your question.
* Note, what I said here isn't actually true. The electron would never make it back to the power company, and is unlikely to even leave the chip -- electrons don't move that fast. But a current would be set up that goes all the way back to the power company, and a bunch of electrons along the path would move a little bit, like a sub-atomic conga line, so the loss of power is the same.