On ARM Cortex-A, I see a list of ARM Cortex-A series CPUs. I want to check each Cortex-Axx CPU's speed (xx Hz as far as I know), so I can know which Cortex-Axx is faster than another. But there aren’t any speed specification for these Cortex-A CPUs.

After googling around, I think I might get something wrong. I guess it's the manufacturer decides the definite speed of an actual Cortex-Axx CPU.

I am lacking background knowledge. What is the explanation?

  • 9
    That aside, judging "a CPU is faster than another" by looking at its clock frequency is often misleading, as its internal architecture design weighs far more than the clock frequency.
    – iBug
    Feb 8 at 7:14
  • 2
    Megqhertz Myth
    – gronostaj
    Feb 8 at 13:26
  • 2
    Don't think of "ARM Cortex A-72" as the name of a specific chip. ARM doesn't build chips. What they created is a great big block of VHDL code (or something equivalent), as well as a spec describing how a "Cortex A-72" core works. Any chipmaker who pays them enough money can have a copy of that code, and then they can make chips that behave as CPUs according to that spec. The spec doesn't specify a speed, so the speed can be whatever the chipmaker or user chooses, as long as they can make it work. Feb 9 at 3:00
  • 1
    Just knowing the max MHz wouldn't tell you which is faster, at least not a good estimate of the performance ratio. e.g. a Cortex A76 (en.wikipedia.org/wiki/ARM_Cortex-A76) can fetch and execute 4 instructions per clock cycle (best case), vs. Cortex A55 being 2-wide in-order. en.wikipedia.org/wiki/Comparison_of_ARM_processors#ARMv8-A . A fancier core can get much more work done per clock cycle. (See Modern Microprocessors A 90-Minute Guide!). In big.LITTLE setups, usually the big cores can clock higher, too. Feb 10 at 1:33

1 Answer 1


Because ARM processor core designs are licenced out and then made by many manufacturers who each attach different GPUs and other peripherals in the silicon, along with different optimisations to the core itself for particular needs.

Then each manufacturer may be using different manufacturing processes, one may use a smaller feature size which affects power requirements and could allow a higher clock speed in certain areas.

As a result ARM as a company can only list a speed based on a very specific core, on a specific process, with specific peripherals. That combination may never actually be seen in the wild. Generic clocks speeds are no longer relevant for end users.

If you need to know the speed of an ARM processor then the core version such as A53 or A78 isn't going to help much, you will have to look at specific end processors that are manufactured such as the Exynos 2100

  • 11
    No, it means you don't search a speed rating of a core, but the rating of concrete devices implementing those cores in silicon. AMD and Intel is the same, by the way; there is no speed rating for the "Zen 3" or "Ivy Bridge", there are ratings for CPUs that implement that core. The only difference here is that in Intel and AMD case the design of the core and the production of the device is carried out by the same company, but in case of ARM those are different companies. Feb 7 at 11:36
  • 7
    @Peregrino69 no, i meant different silicon feature sizes. For example one processor may be manufactured at a 7nm (7 nanometer) production while another might use the 7nm+ production process with improved performance. Same essential core features, but better silicon manufacturing processes can potentially change what a speed a CPU can get up to. techcenturion.com/7nm-10nm-14nm-fabrication while ARM might give an example node to target for a core it is down to the manufacturer what they finally decide to do based on cost vs performance.
    – Mokubai
    Feb 7 at 15:23
  • 3
    Exactly this. The ‘Cortex’ names are effectively microarchitecture names in x86 terms, not processor models. Feb 7 at 21:56
  • 2
    @AustinHemmelgarn: Yes, Cortex-A76 or whatever is a microarchitecture, so for example the same loop will run at the same clocks per iteration however it's manufactured. But unlike x86, the microarchitecture isn't tied to a specific process size. With Intel for example, Broadwell is mostly just a die-shrink of Haswell, but had some microarchitectural changes. I guess a good example is that Kaby Lake cores are microarchitecturally identical to Skylake, with a slightly tweaked manufacturing process (14nm+). Anyway, TL:DR: the word "microarchitecture" isn't specific to x86. Feb 8 at 5:15
  • 2
    @ScottishTapWater: For any given part, though, there's a max clock speed it's rated to work reliably with. Clocking lower often makes sense for embedded, to save power when you don't need more performance, but it is meaningful to rate ARM chips with a single MHz or GHz number. (Or multiple numbers for different voltages...) A whole chip will have one or more cores of some microarchitecture(s), often in a big.Little arrangement for high-end chips with some high-performance cores. Feb 9 at 3:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.