Processor power consumption question, error in my calculations?

[closeratio]

This is more of something that came to me rather than a problem that I have, and I thought it'd be more appropriate in here rather than the general hardware subforum.

Ok so I was just looking though my BIOS the other day chacking that my voltages were right and stuff (mainly because I was bored ), and I noticed that I was butting about 1.45v through my cpu (overclocked to 3.5 from 2). Now i didn't think about it at the time, but whilst I was again bored, I suddenly started thinkin about the relationship between power, voltage, and current.

So for example, because it's the only processor with the power consumption I know of, take the Intel Q6600 (the G0 stepping one), which consumes 95W. If the stock volts on that are approximately 1.3v, and the power rating at stock is also 95W, does that not mean that there's something like 73 Amps going through the thing?? (because P=IV, so I=P/V) Because that struck me as a lot, and I know they get hot, but to have that much current going through them they'd surely melt!? (or explode or do something dramatic like that )

Is there something I don't know about the voltage, like is that 1.3v going through each of the little pins on the mobo, meaning that that's 73amps divided by how many little pins there are?

Just struck me as odd, because I don't like the idea of 30+amps going through my CPU lol (E2180)

Cheers guys

 

[theheyes]

Good question, and I hope someone comes up with a good answer. All I know is that the 95w figure is the worst cast scenario. When the CPU is not under full load it will draw a lot less than that.

P.S. Is that Hyde in Manchester?

 

[closeratio]

P.S. Is that Hyde in Manchester?

Yeaha. Why you live nearby or somethin?

 

[Amonlym]

yes, modern cpu's really do draw huge amps!
http://www.benchtest.com/calc.html for working out your overclocked wattage.

i have had in excess of 110 amps (calculated) on cpu in sig. i think this is 1 of the reasons for the forrest of pins and all the gold on cpu's - if there are 30 pins, say, carrying the power thats only 3 2/3 amps / pin. not so horrific (yes, the power is split accross multiple pins)

 

[neodude]

Your corrections are correct. It is this large current going through the CPU that causes all the heat, hence the need for massive heatsinks. As pointed out this current is usually split between a number of pins (Usually around 1 amp per pin). Around 80% of the pins on a Core2Duo are Vcc or Vss pins.

 

[V|per]

Your corrections are correct. It is this large current going through the CPU that causes all the heat, hence the need for massive heatsinks. As pointed out this current is usually split between a number of pins (Usually around 1 amp per pin). Around 80% of the pins on a Core2Duo are Vcc or Vss pins.

There are pins on a core 2 duo now?!!??!

 

[neodude]

Pins/Contacts/Lands/Pads. Whatever.

 

[closeratio]

Lol Funny story actually when I was puttin my current spec together, had sort of a shock when I saw the CPU had no pins thought i'd bent them all lol. Had a look at the mobo and realised they do it the other way around nowadays.

About the cpus though, if an overclocked cpu takes around 130watts, that's like as much as or more than a high end graphics card, which is sorta scary when you always hear people asking about if their PSU will be able to cope with the graphics card they're thinking of getitng

Maybe they should take the CPU into consideration as well

 

[crazyDAJT]

but remember when you overclock the processor will get less efficient from a energy point of view. so should draw more amps than you would expect.

 

[closeratio]

but remember when you overclock the processor will get less efficient from a energy point of view. so should draw more amps than you would expect.

So does that mean that those 70+ amps are all going down that seperate 4/8 pin cable? It just seems a bit thin for somethin that is meant to take so much current

 

[Lightnix]

So does that mean that those 70+ amps are all going down that seperate 4/8 pin cable? It just seems a bit thin for somethin that is meant to take so much current

Actually, the current is only going down the yellow cables. The black ones are ground.

 

[closeratio]

Actually, the current is only going down the yellow cables. The black ones are ground.

They must be some pretty expensive low resistance cables lol. I guess that's why you try to spend as much money as possible on your psu though... more money = higher quality

 

[crazyDAJT]

They must be some pretty (expensive) low resistance cables lol. I guess that's why you try to spend as much money as possible on your psu though... more money = higher quality

but the voltage is so low, the wattage is therefore so low, therefore the heat to be dissipated is minimal.

 

[muon]

You cant simply use P = VI since I also increases as you increase V in processors

Sub in I = V / R

You get I = (V^2)/R conrolling for resisitance (i.e. processor). So as you increase voltage by x%, you get a (1 + x)^2 increase.

NB: I could be completely wrong as i dont claim to hold any knowledge beyond an A-level in physics.

edit: as for the actual question, 30 Amps by itself says nothing though. You have to look at the energy transferred per second, i.e. the wattage and if the HSF is upto the job there shouldnt be a problem. I think what you are talking about is how a thin wire would melt with a high current going through it, because of a high resisitance. But the resistance is specific to the processor and the actual fab technology used. The pins are nothing in comparison to the actual chip which has 45nm "wires".

 

[neodude]

Actually, the current is only going down the yellow cables. The black ones are ground.

So where does the current go? I'm sure it was primary school science that taught us about electrical "circuits". The ground wires carry exactly the same current as the live wires.

You cant simply use P = VI

Yes you can. I=P/V will give is the exact current being drawn from CPU as long as we can accurately measure the power consumption and voltage supplied.

the actual chip which has 45nm "wires"

Incorrect. The 45nm signifies the oxide insulating layer between the gate of any particular transistor and the substrate. The interconnecting wires are a lot larger and the power suppliy rail for the transistors are larger still with smaller hairline conductors branching off from this feeding smaller chains of transistors.

 

[Corasik]

So does that mean that those 70+ amps are all going down that seperate 4/8 pin cable? It just seems a bit thin for somethin that is meant to take so much current

No, the Yellow/Black cable is 12Volts, the voltage regulators that drop it down to 1.3V are on the motherboard. 95W at 12V is only 7.9 amps. There is likely some 'loss of efficiency' on the regulators, but even so, there are enough cables to carry the amperage.

However, these high AMP circuits are exactly why 3phase, 4phase, and even 6phase regulator circuits were added to motherboard at around the Prescott era. Before that simply 2 phase designs were common place.

 

[muon]

Yes you can. I=P/V will give is the exact current being drawn from CPU as long as we can accurately measure the power consumption and voltage supplied.

I thought the OP was asking how power is affected with voltage increases, something that equation will not tell you correctly for CPUs.

On the other thing about 45nm wires, fair enough. I was wondering why is it that the voltage has to be reduced as you shrink the technology?

 

[neodude]

I thought the OP was asking how power is affected with voltage increases, something that equation will not tell you correctly for CPUs.

On the other thing about 45nm wires, fair enough. I was wondering why is it that the voltage has to be reduced as you shrink the technology?

Fair enough on the first point. As the voltage increases the current will indeed increase and the power consumption will increase in relation to both.

Although the power lines in the CPU are not at 45nm they do get smaller with new technologies, hence they need for the voltage drop.

 

[alexthecheese]

They can't be using 70 or 30 amps, because the fuse on your plug is only likely to be 13amps... surely...

 

[Amonlym]

13A @ 240V = 3120Watt
psu converts this into, say, 10A @ 12v going to the cpu = 120Watt at 80% efficiency = 150Watt from the wall for the 120w to the cpu voltage regs. this 120w @12v then gets converted into say 72A @ 1.5V (with another 12w lost to 95% efficient conversion) = 108W

so 108W cpu = 0.625A at the mains. even a 3 amp fuse (720W) would be fine upto about 550w psu, even one heavily loaded. (and even if i thrash my pc i cant get past about 400W from the wall)