How can my CPU be running faster than the maximum speed?


A customer observed that according to Task Manager, their CPU was running faster than the maximum speed. How is that even possible?

Utilization Speed
66% 2.90GHz
Maximum speed: 2.50GHz
Sockets: 1
Cores: 2

The processor's maximum speed is 2.50GHz, but it's running at 2.90GHz? I know that athletes give 110% effort, but this processor is giving 116%!

This is a feature of Intel Turbo Boost Technology 2.0, which runs a processor above its rated frequency if they are "operating below power, current, and temperature specification limits."

To me, this sounds like the CPU can dynamically overclock itself.

Comments (34)
  1. Ray Koopa says:

    Shouldn’t “Maximum Speed” get adjusted to match “Speed” then, or at least renamed to something along “Default Speed”? =3

    1. Ben says:

      Well, Max Speed is the speed it can sustain on all 6 cores. The Turbo speed can’t run if all cores are active because the current and heat are too great.

      The alternative is to say Max Speed is 2.9GHz, but we throttle it down to 2.5GHz for 95% of the time, because of heat/current.

      That would be misleading. Better to say 2.5GHz, plus sometimes a bit more.

      1. ajanata says:

        Not entirely accurate. My 3930k says max speed is 3.6 (or so) but it’ll happily hum along with all 6 cores at 4.2 just fine. It could go even higher if I actually tried to tweak settings.

      2. More complicated than that – the Max Speed is the speed it can sustain indefinitely on all cores given a system build that meets Intel’s minimum specifications for CPU voltage regulation (droop time etc) and cooling. If your system cannot sustain Max Speed, it’s faulty.

        It is, however, possible for a CPU to maintain Turbo indefinitely on all cores, if your voltage regulation and cooling are up to the job. Voltage regulation is a function of motherboard design and of PSU, cooling is a function of the case setup (fans, airflow routes etc). If your motherboard designer has done a good enough job of coping with load transients, and your PSU is giving a stable enough 12V, and the cooling system can keep the upper surface of the chip at a steady 20 °C, then you can Turbo forever on all cores. I’ve seen this demonstrated by Intel themselves (power fed via a double-conversion UPS to the PSU, to stop mains transients disturbing the power supply).

        Put ever so slightly differently; Turbo is the speed the CPU is designed to max out at, and in the Olden Days, would have been the CPU clock speed. Max Speed is a deliberate underclock by Intel, to allow for the fact that it’s become very hard to build an efficient system that takes in power from the mains, provides the stable power the CPU needs, and runs in warm environments. It is, however, easy (or rather no harder than it’s been since Pentium II days) to build a system that runs at Max Speed all the time, and then boosts to Turbo speed whenever it’s able to.

        1. DWalker says:

          A steady 20 C? 68 F? That’s pretty cool.

    2. Ken in NH says:

      Under promise and over deliver.

  2. DWalker says:

    Yep. Some processors have a rated “max speed” and also a rated “turbo boost” speed. This is useful if the OS, the motherboard, the BIOS, the northbridge chipset, and the CPU all cooperate.

    So, yes, “max” is a misnomer, but oh well.

  3. wqw says:

    Maximum speed does not reflect any overclocking that was setup in BIOS too. It’s the “stock max. speed” as reported by CPU.

  4. Piotr Siódmak says:

    I noticed CPU usage above 100% (even 1000%) in resource monitor. It happens when resmon can’t gather samples – there’s a dip in the graphs and when it comes back it reports over 100% usage for a moment. Is it also due to dynamic overclocking or a side effect in resmon (or ETW)?

  5. Steinar H. Gunderson says:

    Yes, in a sense, it’s an overclock.

    What’s not obvious is that it only applies when you’re not using all the cores. CPUs could generally be faster than they are today if we had better ways of dealing with the power/thermal issues (ie., the CPUs live just fine on the higher clock, just not for very long). But when most of the cores are idle, you have more thermal headroom and can boost the clock frequency a bit. If you look at CPU specs, you can sometimes see Turbo Boost specified as something like 2/2/3/5/etc., which says how many multiplier steps they are willing to add for various numbers of cores in use.

    Also, some features tend to inhibit or reduce Turbo Boost; in particular, using AVX heavily. There’s a lot of rather complicated circuitry in the CPU to try to measure what’s going on and how much power that is needed (it’s not necessarily directly measured).

    1. Brian_EE says:

      >Yes, in a sense, it’s an overclock.

      In what sense? Not a technical sense. Technically it is not an overclock. Overclocking is running the processor clock beyond it’s manufacturer stated limits. In this case, the manufacturer has said that under certain operating conditions, then the clock can run faster. You are still within the processor’s stated specifications.

      We have long had dynamic clock scaling in the embedded world for power management. In that case CPU clock frequency is usually dependent upon core voltage so you can ramp down clock and voltage and save power in idle/sleep modes.

      1. Sockatume says:

        If we’re going to argue terminology, overclocking is any instance where a CPU is made to run beyond the rated speed on the packaging, even if the new speed is well within the manufacturer’s stated limits for that product. Binning means that a “budget” version of the same CPU often has a lot of headroom.

        But obviously we all know what we all mean and we shouldn’t be arguing petty semantics.

        1. smf says:

          Ancient history lesson. When Motorola started the 68000P10 design (the bit after the P usually denotes the maximum clock speed) they were pretty conservative as they found that all parts would run faster. So the official speed rating is 12mhz, but will run faster than that.

          The “maximum” now is the lowest maximum the cpu should be able to run at.

  6. Adrian says:

    It’s like those signs along the side of the road that say, “Maximum Speed 65 MPH.”

    1. BZ says:

      Except that they don’t. If you see “Maximum” you are in Canada and speeds are in KPH. If it’s MPH, you’re in the US and it says “Speed Limit”

      1. Nick says:

        The signs Adrian is talking about have speakers that audibly state the speed limit for blind drivers, but you probably couldn’t hear it over your engine.

        1. Yukkuri says:

          I am sure glad I wasn’t drinking anything before reading this comment.

        2. JanH says:

          They already had that idea in an old ISIRTA sketch, although there the road signs were supposed to be transmitted into the car’s hi-fi system.

      2. ajanata says:

        There are definitely signs in California that say “maximum speed” and are listed in mph.

      3. Neil says:

        If only signs didn’t need any words then you wouldn’t have to bikeshed over what the wording is. Perhaps you could just paint a red circle around the number or something?

        1. smf says:

          >If only signs didn’t need any words then you wouldn’t have to bikeshed over what the wording is.
          >Perhaps you could just paint a red circle around the number or something?

          If only we did that in the UK. We pretend that we do that, but there are exceptions.

          If there are three or more street lights no more than 183 meters apart then you have to assume it’s 30 mile an hour speed limit, to “avoid” confusion a 30 mile an hour sign should not ever be displayed as they think you’ll get used to relying on the signs and not measuring the distance between the lights. They say that the lighting doesn’t have to be provided by street lamps to count, so you have to guess whether someone is trying to catch you out when you are driving in daylight (at night you would see the lights).

          https://en.wikipedia.org/wiki/Built-up_area_(Highway_Code)

          If you are driving along the road and you see a white circle with a black line across it then you have entered the insane world of the national speed limit. At one point it meant you were allowed to drive at whatever speed you liked, but when there was a fuel shortage they imposed a limit. What type of vehicle you are driving, whether you are towing something and whether there is a barrier between you and the traffic travelling in the opposite direction determines what speed you should be travelling at. There are roads in the UK that have multiple lanes in both directions but somewhat randomly as you drive along you (hopefully) notice that the barrier that was there a second ago is no longer there. If you don’t notice that the barrier is no longer there and slow down then the speed camera they have positioned at the end of the barrier will photograph you and they will send you a letter demanding money.

          https://www.gov.uk/speed-limits

      4. smf says:

        A sign that says “Speed Limit” are lying, there is no limit to the speed. Only a legal consequence if you go faster.

        1. French Guy says:

          Technically, there is, but not human-made vehicle can even approach the speed of light.

          1. Jolyon Direnko-Smith says:

            I don’t know about you but my human-made vehicle _approaches_ the speed of light every time I apply pressure to the Go Faster pedal.

            However, it does of course approach from a very, very long way away. :)

    2. Jolyon Direnko-Smith says:

      I think he may be referring to signs that serve the same purpose as ones here in New Zealand. That is, that while speed limits posted in multiples of 10 indicate legal limits, those posted to an intermediate 5 kph represent maximum ADVISED speeds.

      So on a road with a legal limit of 100 kph you will frequently come to bends with a posted 55 kph speed. Sure, you can legally try to take that bend at up to 100 kph, but you are being advised that anything over 55 kph and you may well find yourself in the field next to the road. Or the ditch. Or down in the valley in a crumpled, burning wreck.

      As well as being visually distinct (black on yellow [advised] rather than black on white with a red ring [legal limit]) the use of numbers ending in “5” is deliberately designed to prompt recognition that this is not a legal limit, but an advisory based on the nature of the road ahead.

  7. Sockatume says:

    IME it’s primarily used to allow a modern multi-core CPU to perform single-core tasks with high performance. If the other cores are essentially idle and producing little heat, you can run up one processor and still stay within your design limits for power consumption and heat production. (Which is tautological, I know.)

  8. Erik F says:

    This is an honest question: does a CPU’s clock speed mean anything in particular when you have hyperthreading, pipelining and all the other microcode optimizations that take place nowadays? Even if you had a 1THz processor, isn’t it going to end up getting bottlenecked by the GPU, memory, and disk I/O?

    I’m glad that the CPU is doing dynamic speed adjustments, but for me the days of ricing my motherboard are long gone: I can’t see any significant performance benefits in doing so!

    1. Scarlet Manuka says:

      It will of course be *especially* if you have a 1 THz CPU that you will see it throttled by the other subsystems. The faster the CPU, the more likely this is to be the case.

      In general, the answer is “it depends on what you’re doing”. But it’s certainly true that for most common tasks you’re far less likely now to be CPU-bound than used to be the case in previous decades.

    2. smf says:

      >This is an honest question: does a CPU’s clock speed mean anything in particular when you have hyperthreading, pipelining and all the other microcode optimizations that take place nowadays?

      Yes, very much so. Pipelinging has been around for thirty years, hyperthreading for fourteen years.

      There would be many challenges in building a 1thz cpu today, but ram speed has lagged behind cpu’s for many years and larger on cpu caches have been quite effective to that particular problem.

      It is more a question of how much that cpu would be worth to you as it would be insanely expensive. If you have an algorithm that would benefit from being run at 1thz and you could monetize that then yes it would be worth it, if you’re just going to chat to your friends on facebook then the answer would very likely be no.

      Which is why there are atom, pentium, celeron, i3, i5 & i7.

      If intel could produce a 1thz cpu that was cost effective then they would do it in a heart beat.

      1. Voo says:

        With the current technology it would just be plain impossible to build anything approaching 1THz CPUs.

        Remember: the faster you make the transistors switch, the more voltage they need and that sadly scales quadratic. The power you’d dissipate for a 1THz would put you somewhere in the “smaller star” category – perfectly unfeasible.

        And that’s ignoring electron migration and all those other fun effects that mean you couldn’t even push that much voltage to begin with – the chip couldn’t handle it.

  9. The_Assimilator says:

    “Maximum speed when loaded to 100% on all cores” wouldn’t have fitted the dialog, I fear.

  10. GEO255 says:

    Can I wire it up to my “Turbo” button on my old Pc case?

  11. Rosyna says:

    It seems like this may also be for marketing purposes? That is, CPU makers set a 100% baseline at the default clock instead of the max possible clock so it shows 100% normally. If they set it to the real max clock, users might be angry?

  12. Phylyp says:

    Furthermore, if Hyper-V is enabled, then the CPU speed reported is always the maximum speed, and it doesn’t reflect the dynamic speed at that point in time. This happens even when you’re not running a VM, because technically the host OS is itself atop Hyper-V.

Comments are closed.

Skip to main content