Conclusion: TDP is Not Fit For Purpose

In years gone by, processors were sold with a single frequency and power rating. It was very quickly realized that if a processor could effectively go to sleep, using either lower voltage or lower frequency (or both) then a lot of idle power could be saved. Going the other way, processor designers realized that for temporary short bursts, a core could run at a higher frequency before it reached a thermal limit. Also, using a multi-core processor meant that either the power budget could be shared across all the cores, or it could be focused in one.

Both AMD and Intel have noticed this over time, and both companies have different attitudes on how they report numbers relating to ‘base frequency’ and related power as well as the bursty ‘turbo frequency’ and related power. Out of those four metrics, the only one Intel doesn’t provide is turbo power, because from their perspective it is system dependent.

(0-0) Peak Power

Intel lets motherboard manufacturers determine how long a system can turbo for, and what that budget is. Intel encourages motherboard manufacturers to over-engineer the motherboards, not only for overclocking, but for non-overclockable CPUs to get the best performance for longer. This really messes up what the ‘default out-of-the-box performance’ should be if different motherboards give different values. The trend lately is that enthusiast motherboards enable an unlimited turbo budget, and the user building their system just has to deal with it.

This means that users who buy the Core i7-10700 in this review, despite the 65 W rating on the box, will have to cater for a system that will not only peak around 215 W, but sustain that 215 W during any extended high-performance load, such as rendering or compute. We really wished Intel put this 215 W value on the box to help end-users determine their cooling, as without sufficient guidance, users could be hitting thermal limits without even knowing why. At this point, 'Intel Recommended Values' for turbo time and budget mean nothing outside of Intel's own OEM partners building commercial systems.

Core i7-10700 vs Core i7-10700K Performance

In the review we highlighted that these two processors have a peak turbo frequency difference of 300 MHz and an all-core turbo frequency difference of 100 MHz. The fact that one is rated at 65 W and the other is rated at 125 W is inconsequential here, given that most end-user motherboards will simply enable turbo all the time. This means the performance in most of our tests between the two is practically identical, and consummate to a 100-300 MHz frequency difference.

In practically all of our tests, the Core i7-10700K is ahead by a super slim margin. At $387 for the 10700K compared to $335 for the 10700, the performance difference is not enough to warrant the $52 price difference between the two. Performance per dollar sides mostly with the Core i7-10700, although users getting the i7-10700K will likely look towards overclocking their processor to get the most out of it – that ultimately is what to pay for.

The other comparison point is with the Ryzen 5 5600X, which has two fewer cores but costs $299. In practically every test, the increased IPC of the Ryzen over Intel means that it sits identical with the Core i7 processors, AMD is cheaper on list price, and at a much lower power (AMD will peak around 76 W, compared to 215 W). AM4 motherboards are also abundant, while corresponding Intel motherboards are still expensive. The problem here however is that AMD is having such high demand for its product lines right now that finding one in stock might be difficult, and it probably won’t be at its recommended price.

Users in this price bracket have a tough choice – the more efficient AMD processor that might be in stock, compared to the Intel processor that will be in stock but more cooling will likely be required.

Gaming Tests: Strange Brigade
Comments Locked

210 Comments

View All Comments

  • Oxford Guy - Monday, January 25, 2021 - link

    'What if I don't agree about the location of the "sweet spot" based on RAM prices local to me, or my own performance needs?'

    Price vs. performance is a different topic. Nice attempt to change the subject, though.
  • Spunjji - Wednesday, January 27, 2021 - link

    It's not at all an attempt to change the subject. The point is that the issue of an "optimal" RAM speed is a moving target dependent on multiple variables.
  • Makste - Saturday, January 23, 2021 - link

    As a part of his audience, I am actually satisfied with this review, as I am looking for the out of the box experience. I'm most likely not to tamper with the manufacturer's recommendations. So he tested the CPUs following both Intel and AMD recommendations. Something I don't see a problem with.
  • Oxford Guy - Sunday, January 24, 2021 - link

    Turning on XMP is hardly tampering, particularly since motherboard vendors provide lists of recommended RAM to use those XMP profiles with.
  • theqnology - Monday, January 25, 2021 - link

    If it shouldn't be the case, shouldn't consumers take it up with Intel, and not with the reviewer? Why does Intel put lower max than the current JEDEC standards (e.g., Intel at 2933 while AMD at 3200)? I ask, if Intel was so confident about their products, why not up the official support?

    This is really the crux of it. Intel wants to make it part of the difference between their Z boards and their non-Z boards, which adds more cost for the consumers, while at the same time washing themselves from this responsibility should something not work (you cannot RMA based on non-JEDEC compatibility). It is also fascinating that consumers think this is on the reviewers.
  • Oxford Guy - Monday, January 25, 2021 - link

    Sure, as long as the reviewer is consistent with the logic.

    That means no reviewing boards that violate the official base clock and turbo behavior.

    Things like that.
  • Spunjji - Wednesday, January 27, 2021 - link

    Then he can't review any board at all.

    Honestly, this "everything must be treated exactly the same otherwise nothing is fair" rhetoric is ghastly and corrosive in whatever domain it's applied.
  • Qasar - Thursday, January 28, 2021 - link

    " That means no reviewing boards that violate the official base clock and turbo behavior. "
    considering that intel doesnt make board makers conform to anything as far as what " default " would be, good luck with this.
  • quiq - Sunday, January 24, 2021 - link

    i want to view the 1070 whit stock box and a h410/470 motherboard thats the real user for a non k cpu

    vs 10700k and z490 aftermarket HeatSink or WC Oc memories etc Real entusiast cpu
  • IanCutress - Saturday, January 23, 2021 - link

    Please go ahead and enforce both Intel and AMD to rate their memory controllers faster then. Its not opinion, it's the literal standard,and the only way to ensure consistency for comparisons across generations. I'm not going to offer one CPU a higher dram overclock than another, that isn't fair, just in the same way I'm not going to overclocked the cores. I regularly dive into AT's audience metrics, and have done for years. If you want data that's different, then please by all means either do your own testing or find other reviews. But you know, also get them to deep dive into microarchitecture as well as get all the behind the curtain info. Also, all our content is free at the point of use. By your logic, I'm also a consumer of content at AT, thus I also pay my salary. Please enjoy.

Log in

Don't have an account? Sign up now