Question Asus 3080 Strix pulling 50 Watt at idle with a single monitor connected

[DonHansDampf]

Hey everyone,

I watched the market for quite some time and found a good deal for a Asus 3080 (non Ti) Strix GPU and bought it. I have and old (4 or 5 years) Corsair 750x PSU, but I only run an i5-8600k and I intended to undervolt the GPU anyway.

Yes I know, undervolting and OC card. But the price was good, I had some good experience with ASUS cards in the past and I always had the feeling that OC cards like that have better cooling with results in a quieter card.

Anyway I installed the card, connected three seperate PCIe cables to it, run a benchmark with the stock (OC) settings. Everything was good.

Started my undervolt and found the sweet spot with 825mV @ 1800 Mhz (if I need more power in the future I still can increase the clock). Everything was stable, everything could be great.

BUT while monitoring the GPU I noticed the power draw while idling: 50 watt. That's way to high. I checked reviews beforehand and users mentioned like ~20 watts or lower while idling.

So I started to check things:

• I only have one 1440p @ 120Hz monitor connected. Changing it to 1080p @ 60Hz saves like ~8 watts which still results in 42 watts (also I wouldn't run my monitor at this setting...)
• In the global nvidia settings the power usage is set to Normal not maximum performance.
• Windows Power Plan is set to Balanced.
• I tried the two different BIOS (Q Mode and P Mode) without any difference (besides probably a different fan curve).
• I uninstalled the driver with DDU, and reinstalled everything (restore default options checked) without any difference.
• I tried to update the GPU BIOS, but the installer only mentions that my card is already updated.
• Tried different software to read the sensors but all list more or less the same power draw.

Nothing fixed the problem for me. Also check my running processes, but the card also downclocks probably to 210 Mhz Core and 405 Memory.

Are the sensors just broken, or does this card randomly suck 30-40 more watts while idling and no review ever mentioned that?

GPU-Z Screenshot while Idle: https://gpuz.techpowerup.com/22/09/06/vts.png


Afterburner Screenshot while Idle:

https://imgur.com/71yT1Bg

View: https://i.imgur.com/71yT1Bg.png



Afterburner Sensors while Idle:

https://imgur.com/QDChx5K

View: https://i.imgur.com/QDChx5K.png



Afterburner Curve (but I also have the problem with Stock settings):

https://imgur.com/KOHoBs2

View: https://i.imgur.com/KOHoBs2.png



Any help would be great. Thank you!

 

[Ralston18]

This:

"I uninstalled the driver with DDU, and reinstalled everything (restore default options checked) without any difference. "

Noted DDU for the uninstall. Was DDU used for the reinstall as well?

Do another reinstall but completely manual - no third party tools or installers. Download just the basic GPU drivers and no other associated utilities or tools. Could be some default checkbox that adds in other software.

Also look in Task Manager and Resource Monitor for any unexpected and/or unknown apps running in the background.

 

[Lutfij]

Welcome to the forums, newcomer!

Just my thoughts on the matter, not all chips will behave the same, both under overclocking and/or undervolting conditions. Silicon lottery if you will. Yes your numbers are higher(when you're comparing to other undevotls) though are they running an identical setup as yours? Could you parse the links to the source of your 20W number by other users?

 

[Rdslw]

Usually if is 20% difference, then it's run to run variance.
I think you should check if in the source you mentioned, they did not connect GPU to integrated card, as that would make sense.
My old 1060 is running at 40C P8 5W when the monitor is connected to integrated, but it's instantly 21W if I connect the monitor to that HDMI.
Maybe check all slots on GPU or display port if you have cable ?
You might just have some stupid config that gpu needs to keep the random chip powered, because you use 4'th slot instead of 1'st.

not all cards are made equal in branded versions.

 

[DonHansDampf]

Noted DDU for the uninstall. Was DDU used for the reinstall as well?

Do another reinstall but completely manual - no third party tools or installers. Download just the basic GPU drivers and no other associated utilities or tools. Could be some default checkbox that adds in other software.

Also look in Task Manager and Resource Monitor for any unexpected and/or unknown apps running in the background.

I didn't install the driver with DDU, and also installed Nvidia Experience again. Gonna try it again and report back, thanks.

About apps running in the background: Already checked everything and also stopped processes without any effect. With the low clocks I get I would guess that also no program is drawing the power. Otherwise the card wouldn't run at 210 Mhz (or I am wrong?).

Just my thoughts on the matter, not all chips will behave the same, both under overclocking and/or undervolting conditions. Silicon lottery if you will. Yes your numbers are higher(when you're comparing to other undevotls) though are they running an identical setup as yours? Could you parse the links to the source of your 20W number by other users?

I have pretty good results while under load. The undervolt works great and like in many guides/reviews mentioned I am able to save many watts compared to the stock clock.

Only the idle power usage irks me. Could this be also 'bad silicon lottery'?

Igor's Lab for example mentioned that a 3080 Ti only draws 14 watt while idling: https://www.igorslab.de/wp-content/uploads/2021/06/03-Power-Consumption-1.png

So does this chart from Techpowerup: https://tpucdn.com/review/nvidia-geforce-rtx-3080-founders-edition/images/power-idle.png

Usually if is 20% difference, then it's run to run variance.
I think you should check if in the source you mentioned, they did not connect GPU to integrated card, as that would make sense.
My old 1060 is running at 40C P8 5W when the monitor is connected to integrated, but it's instantly 21W if I connect the monitor to that HDMI.
Maybe check all slots on GPU or display port if you have cable ?
You might just have some stupid config that gpu needs to keep the random chip powered, because you use 4'th slot instead of 1'st.

not all cards are made equal in branded versions.

I don't think that tech review sites connected their monitor to the iGPU to measure power usage of GPUs (see the charts linked above). But I am gonna check on this and report back. Also didn't know that different outputs could make a difference. Let's see... Thank you!

 

[DonHansDampf]

Update:

Sadly using different connectors didn't help. Also installing the driver without any tools/gimmicks didn't do any effect. The power draw is the same.

When I boot up with the monitor connected to the iGPU, the power load is 36 W, what would be acceptable. But all these sites list something like '20 watt with monitor connected, 30 watt with two monitors'...

https://gpuz.techpowerup.com/22/09/06/79d.png

 

[DSzymborski]

One unaddressed problem here is that these aren't apples to apples comparisons. The things you're linking to are places that directly measure wattage, rather than the less accurate software-reported wattage. There's a reason sites directly measure rather than just go by software reported wattage.

It would also be useful to compare numbers at load as well, not just at idle, because you want to explore the breadth of the inconsistencies. And while doing this, run everything at stock. You're tainting the data because you're making ad hoc fixes while diagnosing the issue.

 

[DonHansDampf]

One unaddressed problem here is that these aren't apples to apples comparisons. The things you're linking to are places that directly measure wattage, rather than the less accurate software-reported wattage. There's a reason sites directly measure rather than just go by software reported wattage.

I know that there is a difference between directly measuring the wattage and using software based measurement (which is more or less a guess based on the used voltage).

Anyway it's also possible to find many screenshots in the web showing for example GPU-Z on a 3080 and the software based power draw shows 10-30 watts and not 50. For example on this site testing the MSI 3080 Trio: https://www.notebookcheck.com/fileadmin/Notebooks/NVIDIA/MSI_RTX_3080_Gaming_X_Trio/gpuz_2.jpg

I also installed my old card (Zotac 3060 Ti), didn't change any windows or nvidia setting, didn't install a different driver etc. and the card goes down to 13 watts idle. So I guess it can not be any setting or application preventing the idle mode. I also checked the power state of the 3080 and it sits at P8 as expected.

It would also be useful to compare numbers at load as well, not just at idle, because you want to explore the breadth of the inconsistencies. And while doing this, run everything at stock. You're tainting the data because you're making ad hoc fixes while diagnosing the issue.

Just to be sure I reset everything to default (stock settings, no MSI Afterburner or GPU Tweak), went into P-Mode Bios and run a benchmark. Under full load we have the expected 390 Watts. So the GPU is hitting it's default power limit without any issues. Also the 3D Mark score fits this (perfoms slightly better than the average 3080).

 

[DonHansDampf]

Update on that topic:

I know have every combination of settings, driver, bios, ... tested without any difference in power draw.

It also didn't matter what BIOS, Profile, Voltage, ... I used. The idle Power Draw stayed at 50 Watt measured via Sensors. The same system showed only 15 Watt power draw with the 3060 Ti.

To be sure that not only some sensor is broken (even when the power draw from the four lanes PCIe, Pin1, Pin2 and Pin3 adds up to the displayed total power draw) I bought a wattagemeter for the socket.

The whole system draws with the 3060 Ti installed 70 Watt. The same system with the same components and same settings and drivers draws 100-105 Watt with the 3080 installed. We have a delta of 35 Watt. The same delta displayed in HWinfo etc.

The 3080 isn't in a different state than the 3060. The clock speed is the same (210 Mhz). The Power State is also P8. I don't measure peaks, i measure the average wattage over 10 minutes of Idle. The average GPU load is in both scenarios ~2%. The only difference? Temperature (3060 = 30 Celcius vs 3080 = 55 Celcius) and TDP Draw (so power draw).

It's good to know that I am not crazy and the sensors were correct.

Either the GPU is broken (and burning energy somewhere) or ASUS simply did a bad job while designing this card. Either way I opened up a support ticket.

 

[LucasMorrison]

youre definitely not crazy lol im runnin a 3090 and 12700k with about 15 tabs in edge and email open only drawing 20w

 

[Viking2121]

Sheesh, My entire system is idling at around 75 - 90ish watts with plex going and bunch of tabs with dual monitors according to my watt meter, I think 50 watts is way to much at idle with a low hz display let alone a single display.... Not too sure whats going on there.

 

[DonHansDampf]

So far ASUS Support doesn't see any problem with the load... My local Support got the answer from Taiwan, that they don't see any problem:

The higher the performance of the graphics card, the higher the wattage consumption;
the higher the resolution and refresh rate of the monitor, the more wattage will be consumed.

Great.

 

[_dawn_chorus_]

Usually if is 20% difference, then it's run to run variance.
I think you should check if in the source you mentioned, they did not connect GPU to integrated card, as that would make sense.
My old 1060 is running at 40C P8 5W when the monitor is connected to integrated, but it's instantly 21W if I connect the monitor to that HDMI.
Maybe check all slots on GPU or display port if you have cable ?
You might just have some stupid config that gpu needs to keep the random chip powered, because you use 4'th slot instead of 1'st.

not all cards are made equal in branded versions.

I just noticed my 3080FE was idling at 55w/210mhz and found your post trying to find a fix. I have a large TV plugged to the HDMI port and a second 27"monitor plugged into the 2nd DP port on the card, both screens running at 1440p 120hz. The second monitor wasn't even on and is in duplicate mode when it is.
Unplugging the DP cable my idle power drops down to 8-15w!
Plugging the DP cable back in, but to the FIRST DP port my idle power stays about 25w, but occasionally does jump back up to 55w.

Whatever is going on here is beyond my understanding but I thought this was useful info to contribute.