Power Consumption

One of the risk factors in overclocking is driving the processor beyond its ideal point of power and performance. Processors are typically manufactured with a particular sweet spot in mind: the peak efficiency of a processor will be at a particular voltage and particular frequency combination, and any deviation from that mark will result in expending extra energy (usually for better performance).

When Intel first introduced the Skylake family, this efficiency point was a key element to its product portfolio. Some CPUs would test and detect the best efficiency point on POST, making sure that when the system was idle, the least power is drawn. When the CPU is actually running code however, the system raises the frequency and voltage in order to offer performance away from that peak efficiency point. If a user pushes that frequency a lot higher, voltage needs to increase and power consumption rises.

So when overclocking a processor, either one of the newer ones or even an old processor, the user ends up expending more energy for the same workload, albeit to get the workload performed faster as well. For our power testing, we took the peak power consumption values during an all-thread version of POV-Ray, using the CPU internal metrics to record full SoC power.

Power (Package), Full Load

The Core i7-2600K was built on Intel’s 32nm process, while the i7-7700K and i7-9700K were built on variants of Intel’s 14nm process family. These latter two, as shown in the benchmarks in this review, have considerable performance advantages due to microarchitectural, platform, and frequency improvements that the more efficient process node offers. They also have AVX2, which draw a lot of power in our power test.

In our peak power results graph, we see the Core i7-2600K at stock (3.5 GHz all-core) hitting only 88W, while the Core i7-7700K at stock (4.3 GHz all-core) at 95 W. These results are both respectable, however adding the overclock to the 2600K, to hit 4.7 GHz all-core, shows how much extra power is needed. At 116W, the 34% overclock is consuming 31% more power (for 24% more performance) when comparing to the 2600K at stock.

The Core i7-9700K, with eight full cores, goes above and beyond this, drawing 124W at stock. While Intel’s power policy didn’t change between the generations, the way it ended up being interpreted did, as explained in our article here:

Why Intel Processors Draw More Power Than Expected: TDP and Turbo Explained

You can also learn about power control on Intel’s latest CPUs in our original Skylake review:

The Intel Skylake Mobile and Desktop Launch, with Architecture Analysis

Gaming: F1 2018 Analyzing the Results: Impressive and Depressing?
Comments Locked

213 Comments

View All Comments

  • mr_tawan - Sunday, May 12, 2019 - link

    Just upgraded to Core i7 4790 (from i5 4460) late last year. At first I was thinking about upgrading to the shiny Ryzen 7, but overall cost is pretty high considering I have my H97 mainboard with 16GB of memory. I don't want to shell out that much money and getting stuck at older platform, again.

    It does work ok, with the performance around the current gen Core i5 I guess (with less power efficiency). Consider what I paid, I think it's not too bad.
  • just4U - Sunday, May 12, 2019 - link

    A interesting read there Ian. I started to notice a slow down on 2600K class systems a few years ago when I worked on them.. (I hadn't used one since 2014) For me.. If I can notice those slowdowns in real time then it's time to move away from that CPU. The 4790K appears to still be holding up ok but older 3000/2000 chips not so well.
  • crotach - Sunday, May 12, 2019 - link

    Still running 3930k Sandy Bridge.

    Maybe Ryzen 3000 will give me a reason to upgrade.
  • AndrewJacksonZA - Sunday, May 12, 2019 - link

    Best quote out of the entire article:
    "In 2019, the landscape has changed: gamers gonna stream, designers gonna design, scientists gonna simulate, and emulators gonna emulate" :-)

    But seriously though, for me, when I upgraded from a Core2Duo E6750 with 4GB of RAM to an i7-6700 (non-K) with 16GB of RAM, it was simply amazing. I was fully expecting that going from an i7-2600K to an i7-9700K would be similar - and it is for things like compiling but not for things like gaming.

    Thanks for the aricle, Ian! Dig the LAN setup. :-)
  • Targon - Sunday, May 12, 2019 - link

    Why would you test a CPU and use a framerate test from Civilization 6, rather than the turn length benchmark which is a true test of the CPU rather than the GPU? Turn based games SHOULD be there as CPU tests, and only caring about the framerates seems to be wrong.
  • Oxford Guy - Sunday, May 12, 2019 - link

    When your overclock fails in one test you're unstable.

    When it fails in four, as in this article, you're both unstable and laughable.

    "Had issues". "For whatever reason". I will assume this is all intended to be humor.
  • DeltaIO - Monday, May 13, 2019 - link

    Interesting article to read. I've only recently upgraded from my 2600k to the 9700k, even that was begrudgingly as the 2600k itself still works fine, however the motherboard simply decided to give up on me.

    I've got to say though, the difference in the subsystems (NVMe vs SSD makes for some great load times for pretty much everything) as well as other tangible benefits (gaming at higher frame rates) is quite apparent now I have upgraded.

    I would have upgraded far sooner had Intel not chosen to keep changing the sockets, swapping out just a CPU is far simpler than rebuilding the entire system.
  • Tedaz - Monday, May 13, 2019 - link

    Expecting i9-9900K joins the article.
  • Badelhas - Monday, May 13, 2019 - link

    I an still with a 2500K overclocked to 4.8Ghz, 8Gb of DDR3 1600Mhz RAM and, a 850 Evo SSD and a Nvidia 1070. I honestly see no reason to upgrade.
    IAN: All your testing basically demonstrated that there is no real reason that justifies spending 400 bucks for a new CPU, 200 bucks for a new Motherboard and 100 bucks for new DDR4 Ram - This totals 700 dollars. But your conclusion is that we should upgrade?! I dont get it.
  • tmanini - Monday, May 13, 2019 - link

    Go ahead and re-read his "Bottom Line" concluding articles: gives a few specific recommendations where is may and may not be to your advantage. And if you aren't desiring/needing all of the other new bells/whistles that go along with newer boards and architecture, then you are set (he says).
    Seems pretty clear.

Log in

Don't have an account? Sign up now