Question idle GPU power draw

[MrSquid_]

Hi,

I've noticed in HWiNFO64 that my GPU power draw (GTX 1080 Ti) is always very high, even idle. I always assumed that to be normal until I started looking into it online and saw reports of idle power draws below 20 W whereas my GPU never goes below 380 W. See below:

Spoiler: GPU idle instant snapshot

All other parameters seem like fine idle values to me : (40°C, GPU core clock below 200 MHz, Memory clock at 101.3 MHz, core voltage of 0.65V, 20% fan speed (it's a blower style GPU for compact cases, it never runs passively), but the reported draw is close to 400 W according to HWiNFO64, the most part of which coming from the 8-pin power input still according to that software.

Is having an idle power draw this high even possible or is this measurement completely unreliable ? If accurate, is there something I can do about it ?

Spoiler: System specifications

Motherboard: ASRock Fatal1ty X99M Killer Intel i7-5820k (OC) Asus GTX 1080 Ti Turbo 4x Crucial Balistix BL8G32C16U4B 16-18-18-36 8GB@3200 MHz Power Supply: Corsair AX760i (760W) Windows 10 Pro 22H2

Cheers,
MrSquid_

 

[Fix_that_Glitch]

I would use Gpuz and look at sensor tab to see what it tells you. TechPowerUp GPU-Z (v2.51.0) Download | TechPowerUp

 

[MrSquid_]

I would use Gpuz and look at sensor tab to see what it tells you. TechPowerUp GPU-Z (v2.51.0) Download | TechPowerUp

Thank you for you quick reply.

I ran gpu-z for a few minutes without touching my PC. Screenshot below. tldr : It tells pretty much the same story than HWINFO64

Spoiler: GPU-Z results

I checked the NVIDIA BIOS in the advanced tab by curiosity and noticed a maximum power limit of 300W, although I don't know if it refers to the same thing.

Spoiler: NVIDIA BIOS

I tend to trust HWiNFO64 as an overall system monitoring tool (I also use MSI Afterburner), but I do question the reliabity of the measurement itself. I don't know how these values are obtained ; I assume they are reported by the GPU itself.

 

[Fix_that_Glitch]

How many monitors do you have connected?

 

[Fix_that_Glitch]

Try uninstalling the gpu drivers with DDU uninstaller, Display Driver Uninstaller Download version 18.0.5.7 (guru3d.com) then install fresh set of drivers. See if this changes anything.

 

[MrSquid_]

I have 2 monitors. One is 144Hz connected over Display Port and the other one is 75Hz connected over HDMI. I am a victim of the nvidia specific issue where in such setups, GPU clock speed gets stuck at its maximum. For that reason, I lowered my main monitor's refresh rate to 120 Hz which is how I manage to get normal idle clock speeds. However the issue is not related to monitors AFAIK. Me using a dual monitor setup is something relatively new and I did already notice a reported 380 W idle power draw back when I only had one monitor.

This issue is not related to drivers either AFAIK. I've ran DDU several times in the past for unrelated reasons and I've also had to re-start with a completely fresh Windows 10 install because of a failed windows update that left with a completely broken partition scheme a few months ago.
I've noticed a high reported power draw from the get-go with my system but am only bringing this up now because of how expensive electricity is becoming in Europe and, while I've suspected my reported value to be incorrect ever since I came accross reports from others and reviews etc, I just wasn't completely sure about it.

I've been looking up more stuff since I posted earlier today and apparently the ATX standard for the maximum power draw with a PCI-E 8-pin connector is 150W, and 75W for a 6-pin connector. Since my GPU is powered through a 6-pin and an 8-pin connector, it can theoretically draw 225W max (not counting the PCI slot itself), which is in line with tom's hardware review which reported 248.6 W when the GPU is being "tortured"

Spoiler: Tom's Hardware review power consumption

At any rate, it seems all the more unrealistic that my GPU is pulling 380 W from a single 8-pin connector like suggested by all the various monitoring software.

I just attempted to measure the power draw from my whole PC in idle state by comparing the reported draw from my electricity meter with the PC being powered OFF and ON and it appears to be around 60 W (which also accounts for my monitors).

At this point, I'd be interested in knowing where the issue precisely lies, but I don't know where the value comes from... Is it measured and reported by a component in the GPU itself, from the motherboard or maybe even from the PSU ? I wouldn't be too surprised if it came from the motherboard because it's not the first time I've noticed strange values from the MB's sensors. For instance, I have some AUXTIN entries (from 1 to 3) in HWiNFO64 with reported temperatures often close or even above 100°C, which I don't know what to think of, as I don't know what these are nor how the temperature is measured within them (but surely not with a thermometer).

I also wonder about the following related matters : Is it possible that this reporting issue comes from the PCI-E cables from my PSU ? Can it be indicative of a hardware fault ? Can it have any negative impact or can it safely be disregarded ?

I'm mostly being curious and would appreciate any information that might enlighten me on the subject.

 

[Fix_that_Glitch]

What does it show for power draw when being used?

 

[lordmogul]

Looks to me like an issue with the readout.

Even if the card were stuck at high clocks (which it doesn't seem to, it sits at normal 139 MHz) it shouldn't be drawing this much power draw in idle. Maybe 15-20W, but not that much.
GPU-Z not only says that the card pulls 390W but that it draws 380W over the 8-pin alone while everything else is looking like it should be.
It also says the card is sitting at 150% of it's TDP, which shouldn't be possible unless you have overclocked it (including increasing the power limit slider)

The fact that the chip sits at 35°C with fans at 20% at 0.65 V also hints at an issue with the readout. There is no way a card pulling 390W would be that cold with that much power draw unless there is some subzero cooling involved.

Do you have a electricity meter for your wall? If the card really draws 390W, it would be reflected there with >400 W at the wall. Idle would be clearly sub 50W with these components.

 

[MrSquid_]

What does it show for power draw when being used?

Not that high, interestingly. Absolute maximum readout just short of 480 W. With the idle value at around 390 W, that's just a 90 W delta. And that's with me using my GPU overclock profile with +250 MHz clock speed, +350 MHz memory clock, and power limit up to 120% (as far as the slider goes), with 99% GPU usage.
The delta is not high enough, that can't be right either. If there was an offset in the readout, I would've expected a delta of around 200W.
Power draw is the second row from the bottom in MSI afterburner in the pic below (Warning: somewhat graphic content in the game screenshot. I wasn't paying attention, sorry about that).

Spoiler: MSI Afterburner, Cyberpunk 2077

Looks to me like an issue with the readout.

Even if the card were stuck at high clocks (which it doesn't seem to, it sits at normal 139 MHz) it shouldn't be drawing this much power draw in idle. Maybe 15-20W, but not that much.
GPU-Z not only says that the card pulls 390W but that it draws 380W over the 8-pin alone while everything else is looking like it should be.
It also says the card is sitting at 150% of it's TDP, which shouldn't be possible unless you have overclocked it (including increasing the power limit slider)

The fact that the chip sits at 35°C with fans at 20% at 0.65 V also hints at an issue with the readout. There is no way a card pulling 390W would be that cold with that much power draw unless there is some subzero cooling involved.

Do you have a electricity meter for your wall? If the card really draws 390W, it would be reflected there with >400 W at the wall. Idle would be clearly sub 50W with these components.

That's my thought too. I do use overclocking, but not at all times. Only when playing specific games in order to squeeze a few extra frames per second out of my gpu. I switch back to default settings the rest of the time.

I don't have an elictricity meter power socket right on the wall, but I did attempt to measure my whole system's idle power draw from my home's meter, which came out at around 60 W (cf my previous message), and that includes all of my peripherals as well.

It's pretty clear at this point that the readout is incorrect, but I wonder how that can be and if I can / should do something about it.