Are the Haswell low-TDP models more energy efficient than the standard models?

henris

Honorable
Nov 20, 2013
2
0
10,510
When comparing i3-4130 and i3-4130T I ran into debate about the TDP and it's relationship to actual energy efficiency especially under full load.

i3-4130 has a 54W TDP and Passmark score of 4970, i3-4130T has a 35W TDP and Passmark score of 3989. Thus the T-model's TDP is 65% and Passmark is 80% of the standard model's.

Given the above numbers and the assumption that Passmark benchmark will put CPU under full load (=100% usage), I have the following questions:
1. Has anyone measured or other wise have knowledge whether the TDP value anyway reflects the power consumption of the CPU when the usage is at 100%?
2. Will a low-TDP consume less energy than a standard model to perform an otherwise identical computing task?

I haven't been able to find any source/test which would have actually measured the CPU's power consumption or total system consumption and compared low-TDP- and standard models. If there was no efficiency difference it would not make any sense to use low-TDP-models unless you had a very limited heat dissipation capability.
 
TDP isn't a precise rating, you will notice that they are always binned, you don't find a CPU rated at 121 W and the next one up rated at 123W, they would be rated at 125W because there is variance in chip to chip efficiency. In general to create the lower power variants they will pick from the standard chips and find ones that are stable at lower voltage, if you can drop the voltage by 10% you can reduce power consumption to 81% of starting, dropping the clock frequency ~15% from 3.4 to 2.9 also gains you 15% heat reduction. Combined those bring you down to ~70% of original power for the CPU portion, that is how they usually make them work. And since they are producing less heat at the same clock speeds they are more efficient.

You will notice their super long life laptops come with ULV(Ultra low voltage) processors, this is because power is influenced by the square of the voltage used so you get the most gains by being able to reduce voltage while keeping the same clock frequency.
 

henris

Honorable
Nov 20, 2013
2
0
10,510
Thanx for the replies. I forgot to mention that one of things mentioned in the debate was an article from Toms Hardware related to Ivy Bridge processor efficiency:
Core i5-3570K, -3550, -3550S, And -3570T: Ivy Bridge Efficiency

It states that the 3570T model is actually the least energy efficient. The numbers are pretty clear and the overall conclusion is quite crushing on the T (and S) models:
"Why on earth would you want one of those low-power parts then? Only one reason: to cram the performance you do get from them into a smaller form factor. In some environments, a 77 W chip simply doesn’t work."

I was under the very same impression as koreanoverlord and hunter315 but the above mentioned article says the opposite. It would seem logical that the same applied also to Haswell processors but does someone know better?