@Parsifal I told you that about half of the power going in comes out as heat.
You asked, where does the rest go.
I gave you a bunch of possible places including communication, losses and mechanical work of physically moving the transistors which have a mechanical resistance, not all of which becomes heat.
Why is this hard to understand? You know you can't put 100% electrical energy in and get 100% heat out. There will be a difference between those two numbers.
Power draw (watts) = work done (watts) + other losses (watts) + heat losses TDP (watts).
You know TDP can't be power draw.
Also some of your objections are weird for a scientifically minded person.
Well, some of the power is passed on to other components.
This is negligible, since the purpose of these signals is to transmit information and not power, so there isn't a lot of current involved. But more importantly, this will just be turned into waste heat when it reaches those components anyway, so it makes no difference to the total heat the system generates.
Well CPUs do send 12v communications to some components so ... but mostly if the energy arrives at the RAM stick and loses heat there, that's not part of the TDP ... the heat coming off the processor die. You know that. That's heat elsewhere.
Some is lost to electro-magnetic radiation.
Which is a form of heat, albeit not one that needs to be extracted by a cooling system.
No, that's generating a magnetic field which causes resistance of its own as it permeates materials such as your copper heatsink. Pushing something metallic into a magnetic field uses energy. Like wise pushing a field out through metals must therefore also use energy because of newton's second law. Sure some heat, but not all heat. Some is going to be left hand rule force in Newtons.
Some is noise, you can hear a CPU.
Not over a 10W speaker. Also negligible.
Its all negligible. it takes 80,000 watts to power a Nissan Leaf. you are asking me where just 67 watts goes. The average human fart produces around 60 watts of power.
Some is vibration. If a CPU operates at 4GHz ... that's a lot of very shallow amplitude but high frequency oscillation. Imperceptible to you as a human, but guzzling power all the same.
4 GHz waves cannot permeate effectively in air. A 4 GHz sound wave in air has a wavelength of about 80 nm, which is short enough that it spans about 20 nitrogen molecules. You're playing pool with molecules at this point. This means the vibrations are never going to leave the computer, and therefore do not qualify as a form of loss.
Guess what happens to vibrations when they impart energy but don't travel very far as waves. Go on, you can think about this one for a bit.
That pool of moelcules is still being moved around. That's a loss. Billions of transistors, all moving gates, shifting molecules millions of times a second, it adds up.
Electrostatic losses when not all power send to capacitors/transistors reaches them.
Another form of heat...
No. That isn't heat. That's electrons that were stuck to something sodding off and having to be replaced by new electrons. Bit like a balloon you rubbed on your head losing its charge after a while.
These days even quantum tunneling is a loss. Not all your electrons do the thing they were sent to do.
"These days"? Are you aware that quantum tunneling didn't just spawn into existence when it was discovered?
Nevertheless, this is both negligible and irrelevant, since electrons are matter and not energy.
Yes these days as the gates are only 14nm and the electrons can tunnel the gap. Much less a problem at 80nm gates. Its a problem that will only get worse as transistor sizes get smaller.
There are loads. A CPU isn't just made to generate heat.
Actually, that is literally what a CPU is designed to do. It's an entropy machine.
No. No, consider how silly that is. It is a counting machine. It happens like all machines to have losses.
Are you familiar with the Second Law of Thermodynamics? It states that the total entropy of a closed system must always increase. Now consider what a computer is supposed to do. The job of a computer is to take disorganised data -- with high entropy -- and turn it into a form that can be easily accessed by humans -- with low entropy.
Now we have a problem. To do its job, a computer needs to reduce the total entropy in its memory banks. Since the entropy in a closed system cannot decrease, this means that the computer cannot be a closed system, and it must generate more entropy than it removes. The more entropy it generates, the better it performs.
This is the technical reason for the common sense correlation between higher TDP and better performance. When it comes to computers, better performance is more waste heat.
No. I can take a 65W TDP processor and have 8 cores like the Ryzen 1700. I can also take a 65W TDP processor like the ryzen 1400. The former performs better because it has more cores. Not because it has a higher TDP. They have the same TDP. I could take a 7th gen i5 and give it hyperthreading. Still the same TDP but it will perform better.
TDP is the amount of heat it will throw off. Not the power draw. Why is this hard? A hyper threaded i7 guzzles a lot more than a non-threaded i5. They still have the same TDP. they don't have the same power draw or the same performance. The i7 does more work. It needs more power. But its heat loss is comparable because the energy from power draw is used in other ways. It is more efficient, hence the reason it performs better and Intel want another $100 out of you for the privilege.
It sounds like y'all are arguing over different parts of the same thing.
Parsifal, Pete, and Rushy sound like they're saying "TDP is electrical power that's being converted to heat" which is accurate. The CPU won't generate any watts of heat without electrical power and it will need to draw at least the amount of power needed to produce the heat it outputs.
Thork is saying that the CPU is pulling more than the heat generated (which is also true), but where he's wrong is the energy it's pulling isn't stored in the CPU. It just passes through. Like, unless I'm mistaken, a CPU isn't going to store a charge inside itself then release that charge to the motherboard to send a signal to read RAM. It'll simply have a path of electrical charge that leads from the motherboard's BUS through the CPU and out to the RAM.
I would personally ony count the electrical power the CPU uses for computation within itself, not what essentially passes through after a calculation.
Of course, having almost no background in microprocessor architecture and engineering, I could be wrong about the above.
The point is, not all the power drawn in, comes out as heat. Its very simple. So TDP is a lower number than power draw. Your chip always pulls more power than it expels in heat. Kinda obvious. Heat is a loss. The chip is doing useful work as well.
total power = useful work - losses (of which one loss is TDP and must therefore be a lower number than total power drawn)