what's the difference between undervolting and power limiting?

what's the difference between undervolting and power limiting? like, my gpu at default is limited to 100w instead of 120w for some reason.
the clock is also higher than it should be according to amd's website. im not sure if i should undervolt in this case or do something else to optimize my gpu.

A Conspiracy Theorist Is Talking Shirt $21.68

Tip Your Landlord Shirt $21.68

A Conspiracy Theorist Is Talking Shirt $21.68

  1. 1 month ago
    Anonymous
  2. 1 month ago
    Anonymous

    Undervolting means running at lower voltage than stock for some particular clock frequency. Lowering the voltage reduces power consumption (for that particular voltage / frequency point) which also reduces temperature, which in turn means the device might run at higher clock and as such improve performance due to the extra boosting overhead available. The downside is that if you undervolt too much the device will become unstable since the voltage will no longer be enough for it to run correctly.

    Power limiting just sets a power limit, it does pretty much what it says on the tin. The device will run at a point on its stock voltage / frequency curve such that the power it uses is below the configured limit.

    • 1 month ago
      Anonymous

      >The device will run at a point on its stock voltage / frequency curve such that the power it uses is below the configured limit.
      and what exactly is the use case for this?

      • 1 month ago
        Anonymous

        The use case at stock settings is to keep the device operating in its design parameters, it's one of the limits which control clock boosting (others being temperature, the max clock, max voltage, etc.). As to why you would adjust it, maybe you want it to use less power so it's quieter, or you want it to use more power so it runs slightly faster.

        • 1 month ago
          Anonymous

          but what should i do in my situation? like where by default it's power limited to 100w but also clock speed is apparently higher than it should be?
          also im on linux, im not sure if the automatic boost clock speeds are a thing on here.

          • 1 month ago
            Anonymous

            I don't know, depends what problem you're having and what you want to solve I guess? If the card works well, is stable and is performing as it should or better, then you can leave it be.

            If you want to tweak, then undervolting can increase performance at the risk of instability or you can increase the power limit which might improve performance a little bit at the cost of more heat and power consumption.

          • 1 month ago
            Anonymous

            junction temperature reaches 90°. website says my card is safe up to 110° but it feels too high for me.
            also im scared of it being accidentally overclocked with possibility of damaging my hardware since the clock doesn't match what it's supposed to be but it's way higher.

          • 1 month ago
            Anonymous

            I haven't had any AMD card in a long time so I can't comment as to that number, though it's definitely hotter than my NVIDIA cards get. Still if your card is overheating then you have a problem on the cooling side of things and fricking about with clock and voltage settings isn't really the solution, you need to fix your cooling instead.

            Overclocking does not damage hardware, overvolting does.

          • 1 month ago
            Anonymous

            >if your card is overheating
            technically it's supposed to be safe up to 110°, according to amd themselves. but it's reaching 90° and im worried because it still feels too hot and summer is coming as well.

          • 1 month ago
            Anonymous

            20C to the limit doesn't sound too bad, but yeah 90C in general is quite hot. Graphics cards often come with fan curves tuned for silence rather than lower temperatures, so assuming your case airflow is already fine then the simplest and most effective solution to lowering temps is just increasing the fan speed.

          • 1 month ago
            Anonymous

            case airflow is an absolute meme. the fans that come with your case should create more than enough positive pressure to push ambient air outward, unless your shit is caked with a literal blanket of dust or you live in arizona/the sahara desert with no air conditioning.

          • 1 month ago
            Anonymous

            >these unspecified fans which exist in unspecified number (down to 0) and are installed in unspecified positions with unspecified airflow in this unspecified case with unspecified airflow restrictions and an unspecified amount of heat being dissipated by components inside """should""" be more than enough
            Thanks bro, you're very useful.

          • 1 month ago
            Anonymous

            Who are you quoting?

          • 1 month ago
            Anonymous

            airflow with more holes is actually worse
            a lot of laptop "revisions" back when it badly needed fans (when it tried to be like slim apple shit but houses a gayming dgpu) would actually have modification (a thermal plastic) that blocks intake vents just so the air flow is more localized and surprisingly when I tried removing those blocked intake vents the temps went higher so I actually blocked intake areas for my old case, so that it air flow would target the cpu and GPU and also the drives

          • 1 month ago
            Anonymous

            90° is fine but if you're concerned about temps then definitely undervolt and underclock it. Lowering the power limit will kneecap your performance but if you're undervolting you can turn it up without worry: the GPU will still pull less power (and therefore generate less heat) but it will have access to more power when it needs it to maintain its target frequency. If it ever does this it should only appear as a transient spike.

          • 1 month ago
            Anonymous

            90c is below average for junction temperature, are you moronic?
            >but yeah 90C in general is quite hot.
            ...for average temperature, not junction temperature

  3. 1 month ago
    Anonymous

    Undervolting is physical, you are actually lowering the voltage (albeit via software) supplied to the chip, it is a passive process.
    Power limiting is an active digital process requiring feedback and closed control, so the digital circuitry knows when to limit.

    The former may not necessarily be stable, i.e. computations may produce errors, whereas the latter is, it runs at normal voltage and limits when to use the chip.

    At least please tell me this is how they actually do it, my last overclocking was done on a C2Q a decade ago, lol.

  4. 1 month ago
    Anonymous

    >what's the difference between undervolting and power limiting?
    power limit does not interact with volts, you can use both actually. limiting is also not guaranteed to just limit it 100% after a kernel crash, so you want a volt limiter at the ring negative level, not application (proprietary) "kernel" level
    >like, my gpu at default is limited to 100w instead of 120w for some reason.
    haha this and those are two different things. it reporting 100W is more of a loonix thing, based from what looks like yaru theme from your screenshot, it's highly localized as linux issue, even on my nvidia gpu
    >the clock is also higher than it should be according to amd's website. im not sure if i should undervolt in this case or do something else to optimize my gpu.
    >clock higher
    that's like also my ISSUE with loonix, it's even just a toaster gpu, but it is somehow overclocked (clock rate not even on any same gpu variation listed in techpowerup!) though the memory freq is the only accurate one, it's just that GPU specs are whack on linux.
    much like how your clock timings are wrong on linux (e.g, instead of 800-2400MHz it will go like 739-2341MHz etc.) this is actually a bad thing since instead of using the normal rates which have normal voltage stuff it doesn't and this sometimes causes perf regressions like what happens when your cpu just turboboosts+pstate/cstate and suddenly audio goes to shit (don't worry "community" says you just have to input this kernel command telling some shit).

    TL;DR
    >100w instead of 120w for some reason.
    >the clock is also higher than it should be
    loonix moment!

    • 1 month ago
      Anonymous

      >loonix moment!
      yeah i kinda regret loonix but im too lazy to reinstall windows.

  5. 1 month ago
    Anonymous

    Any similar software for AMD GPU's on Linux?

    • 1 month ago
      Anonymous

      that's literally a software for amd gpu on linux. it's corectrl.

      90c is below average for junction temperature, are you moronic?
      >but yeah 90C in general is quite hot.
      ...for average temperature, not junction temperature

      oh yeah whole time i talked about temperatures i was referring to junction temperatures just to be clear.

Your email address will not be published. Required fields are marked *