2

Please ignore this duplicate flag. It's wrong!

Let's say I have the following situation: I have a graphic card which can, with default settings, draw up to 250W from the PSU under full load. However, I don't actually need the maximum performance. I just need the features that come with this card, like a large amount video memory. Thus I will reduce the power limit with software. Most modern GPUs, or at least the GPU I am talking about, do have a power target setting where some sensors inside the graphics card can cap the power usage at this level. You can typically reduce the power by a lot while only sacrificing a bit of performance, as for example reported by this guy.

I know that 250W combined with all other components at maximum power usage will be too much for the PSU. Not by a lot, but still. Let's say it would be totally fine if the GPU drew only 200W from the PSU.

Since I never intend to use the full 250W and will turn the GPU into a <200W card after startup, this should be fine, at least after startup. The only critical phase therefore is the time before the operating system has loaded and applied the software patch that caps the power usage of the card. It's only a temporary fix that will be in place until reboot.

So the question is whether at the moment I turn my PC on (and the software fixes are not yet in place), all components including the GPU will initially draw their maximum power usage from the PSU at the same time and thus will cause the PSU to be overloaded in this instance of time. Or whether the GPU will only draw its maximum power if it is actually busy with a task and thus the PSU will not be overloaded during the startup period.

Please note that I don't want advice in the form of "will this specific PSU be enough for my system", but I want to understand in more detail what challenges the PSU has to deal with, like in the present case about initial startup power usage, so that I can make an educated decision myself.

2

I'm not sure this can be answered precisely without measurements, but here's some generality.

Devices containing motors draw a momentarily high initial current when they start (it's due to the motor's magnetic fields), but then the current rapidly goes down while they run. GPUs don't have that kind of startup effect.

So you're only dealing with what the GPU circuitry is doing. The GPU's power usage is related to the amount of processing work it's doing. There some good discussion on this question.

The GPU's power rating is what it would use at it's maximum processing capability. When you first start up, the GPU isn't doing any intensive processing, so its power usage should be well below the maximum.

You're talking about limiting its maximum power usage through software that loads after startup. That will affect the power it uses when it's actually doing work. But at startup, it isn't doing anything significant. So the fact that the software isn't loaded yet doesn't really make any difference.

  • Thanks for your thought! The reason I was worried is that at startup the GPU may be in its highest clock state (e.g. P0 for Nvidia cards) and thus need the same as if under heavy load. i.e. I feared that full power usage is the "natural" state of the GPU and (idle) power saving is then only achieved by reducing its clock speed or setting some features to sleep. But from your answer I can infer that this is not how it works, right? – manuel Aug 12 '18 at 20:14
  • 1
    The clock state is more like an environment in which it can do its work. A high clock state enables it to process a heavier load, but it is still the actual work that drives the power usage. – fixer1234 Aug 12 '18 at 20:23

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.