i7 920QM vs i7 620M?

Oh, I see, so you were calculating backwards from the 5 watt figure. Hrm. Wait. Now that I'm looking at that chart, I see that they actually list the maximum power as 14.2 mW per output bit at 2.8 GT/s... so I don't know where you're getting the 80 mW figure... unless you're adding in the AC power per output bit, but I don't know how the DC power per output bit interacts with the AC power per output bit. And wait, that power list is per bit, so since we're talking GHz and GB, that's an entire order of magnitude larger... I'm totally confused now.

 

Yeah, I added the AC power to the DC power, since that makes sense as a maximum figure - presumably, actual power consumption could only be less. By the way, Giga- is ~9 orders of magnitude, not one ^_~

Power per bit isn't a meaningful measure, however, because power is energy per unit time - 80mW is 80mJ per second. Energy per bit makes a lot more sense as a measure of efficiency.
Since each link transmits 1 bit at a time, I assumed that the 80mW is per link (because each link would obviously have single-bit width; 16 links = 16-bit bus), which makes a lot more sense. However, a single link at 1.4GHz transmits 2.8Gbit/s in one direction, or 0.35GB/s.

 

That would probably require making a new motherboard with integrated ammeter, not just modifying an existing one. Not unless you built some sort of interface device between the socket and cpu to accurately log and measure the variables.

Temp sensors are totally different. They are actually temperature sensors and the direct readings are displayed. There are not such things as "wattage sensors."

Although CPU voltage is usually known fairly accurately, its current draw is not available, so power consumption is not calculable. With absolutely no way to calculate power consumption, any figure given is just a guess. And by wildly inaccurate, people with i7's at over 4Ghz showing only 140 watt power consumption at load is what I would call "wildly inaccurate." Most of the time these power consumption figures seem plausible enough that they are taken at face value, even though they are arrived at so obliquely and have little merit as relevant data.

Unless you can show me how we can actually arrive at accurate figures for a processor's power consumption, I am going to continue to say it is a shaky guess at best and phony baloney at worst.

 

If the CPU is powered from a separate 12V rail to the rest of the system, it would be possible to measure its consumption separately with relative ease. Otherwise, you would have to go to some effort, like cutting a track on the motherboard and routing it through an ammeter.

 

Well, as lackofcheese said, if you can identify the power line to the CPU socket, it shouldn't be too hard (assuming you're good at soldering) to "break" the line and insert an ammeter. And, as you said, voltage is easy enough to measure. Of course, at this point, we're not really talking about a portable notebook anymore, but you probably would only be putting together something like this as a test rig for CPUs anyway...

 

Haha duh. Sorry, I was being stupid. You could separate the ends of a couple of mosfets from the motherboard and connect an ammeter between them and where they need to go. Shouldn't actually be very hard at all and can be removed. Makes me want to try it on a Dell GM45 socket P motherboard I have laying around.