Unusually high 5070Ti idle power consumption

[Ilstar57]

Recently, I got a new Asus Prime 5070 Ti, but unfortunately it suffers from a problem. Even when only hooked up to a single 1440p 180hz monitor, it idles at 35-40 watts. That's without anything happening, just desktop. Even with the monitor off, it doesn't drop under 30. Fans are off (0 RPM) and no RGB is enabled.

Most reviewers show that 15-20W is the expected idle power for this GPU, and even the most power hungry 70Tis shouldn't go over ~28.

Things I've tried so far:

- Restarting the computer numerous times
- DDU from safe mode x2
- Removing drivers with Revo uninstaller
- Installing latest GameReady drivers (clean installation)
- Switching to latest Studio drivers (clean installation)
- Rolling back to first drivers that support blackwell cards (clean installation)
- Changing Windows power blan to blanaced or power saver
- Changing the power setting to "normal" from nVidia control panel (optimal power is not available)
- Switching between performance and quiet VBIOS
- Undervolting and power limiting with Afterburner and GPU Tweak
- Switting the displayport output
- Switching between 2 and 3 8-pin connectors for the 12-pin adapter
- Updating motherboard BIOS
- Lowering refresh rate on the monitor
- Windows SFC scan


In case you're wondering, my power supply is a 760 watt platinum PSU. With my previous card (GTX 1080), I had about 13-15W of idle power on the same power supply.

Anyone got any ideas on things I could try? Thanks! Also, any other 5070 Ti (or 5070/5080) owners who wish to share their idle power consumption would be more than welcome!

 

[muon]

It is possible at 180hz, that is the best you will get.

Have you tried 60/120/144 hz?

Have you also checked if the clocks are actually downclocking at idle? If not, a program is applying some kind of load.

 

[Ilstar57]

It is possible at 180hz, that is the best you will get.

Have you tried 60/120/144 hz?

Have you also checked if the clocks are actually downclocking at idle? If not, a program is applying some kind of load.

Hey there and thanks for the reply! I tried going down the refresh rates, even the resolutions. At 1080p60, it only drops down maybe ~3-4 watts.


I'm attaching an image which should show what my GPU is doing. Does this appear normal?

 

[muon]

Your GPU chip is throttling down sensibly and ~13W average power draw at idle seems spot on given 2% average load.

I would suggest there is some quirk of the Asus card resulting in everything else using 20W rather than very little.

 

[Ilstar57]

Your GPU chip is throttling down sensibly and ~13W average power draw at idle seems spot on.

I would suggest there is some quirk of the Asus card resulting in everything else using 20W rather than very little.

That's what I've been thinking. However, I've seen other people on the internet and they comment that their power draw with the same Asus card is about 13-17W, which is what's causing me to wonder if there's anything wrong with mine. Yours looks perfectly well. Though the 5070 should draw less power in general than the Ti, it should only be a few watts, not 25. May I ask which 5070 you have?


I also got myself a wall meter. Before plugging it in, I calculated an estimated power draw of all my system components, plus the efficiency loss of the PSU, at around 157W.

Imagine the smirk look on my face then, when upon booting up with the wall meter plugged in, I got 157W exactly. However, soon after that it fell drastically First to 136, then down to about 93-98W power from the wall.

I can't explain it. It's down 60W from where I expected it to be. This is the power consumption I expected of my CPU and GPU alone. Are the sensor readings completely wrong?

 

[muon]

I think your card isn't idling perfectly. There is some load on the card.

Your card voltage should be a solid 0.8000v. I've updated screenshot with system idling perfectly now.

I personally wouldn't worry about it, whilst 35W is high compared to the 20W you might expect, there are too many variables here to be confident in achieving it, short of reinstalling windows.

 

[muon]

I also got myself a wall meter. Before plugging it in, I calculated an estimated power draw of all my system components, plus the efficiency loss of the PSU, at around 157W.

Imagine the smirk look on my face then, when upon booting up with the wall meter plugged in, I got 157W exactly. However, soon after that it fell drastically First to 136, then down to about 93-98W power from the wall.

I can't explain it. It's down 60W from where I expected it to be. This is the power consumption I expected of my CPU and GPU alone. Are the sensor readings completely wrong?

Sub 100W is pretty normal for a system at idle. You are probably overestimating how much components other than CPU/GPU use at idle.

I wouldn't say sensors are wrong.

 

[Ilstar57]

I think your card isn't idling perfectly. There is some load on the card.

Your card voltage should be a solid 0.8000v. I've updated screenshot with system idling perfectly now.

I personally wouldn't worry about it, whilst 35W is high compared to the 20W you might expect, there are too many variables here to be confident in achieving it, short of reinstalling windows.

It's possible. I did try it on another system that had a much more barebones installation of Win 11, and the idle was similar, tho.

Sub 100W is pretty normal for a system at idle. You are probably overestimating how much components other than CPU/GPU use at idle.

I wouldn't say sensors are wrong.

Yeah, I would think so myself. However, my 5700X3D idles anywhere between 35 and 50W, and if we then take another 35W for the GPU, we're already up to 70-85W of component power draw. Assuming a ~10% efficiency loss from the PSU, that puts me at around ~77 - ~93 watts from the wall. Could it be possible, that all my remaining system components, including motherboard, RAM, x3 storage drives, wifi card, CPU cooler, case fans, etc could all suffice with so little power?

 

[Tetras]

Could it be possible, that all my remaining system components, including motherboard, RAM, x3 storage drives, wifi card, CPU cooler, case fans, etc could all suffice with so little power?

Yes. On average, those parts are only a watt or two each (idle). Motherboards vary a lot, anywhere from 5 to 50.

 

[Ilstar57]

Yes. On average, those parts are only a watt or two each (idle). Motherboards vary a lot, anywhere from 5 to 50.

I have an actively cooled X570 chipset-based mobo, so I assumed that would be at least 15-ish. I guess I lucked out on the other components, if not the GPU?

 

[Ilstar57]

PS. The card nearly burned my house down. nVidia and Asus both agreed it was due for an RMA. When I put it under stress test for 20 minutes it melted the PSU. RMAing both the GPU and PSU. What helped lower idle on my older card is turning off "automatically manage colors for apps" from the profile/color management settings of the display tab in windows settings, but this didn't appear to help the 70Ti.

 

[Tetras]

PS. The card nearly burned my house down. nVidia and Asus both agreed it was due for an RMA. When I put it under stress test for 20 minutes it melted the PSU. RMAing both the GPU and PSU.

I'm sorry to hear that.

I'm just curious, what was the PSU model?

How did you have it connected? Did you use the adapter in the box, or a native 12vHPWR/12v-2x6 cable?

Did the melting occur at the cable/connector, or somewhere else?

 

[Ilstar57]

I'm sorry to hear that.

I'm just curious, what was the PSU model?

How did you have it connected? Did you use the adapter in the box, or a native 12vHPWR/12v-2x6 cable?

Did the melting occur at the cable/connector, or somewhere else?

Yeah, sucks, but luckily I was able to get a new PSU in time before I RMA this one, so at least I'm not left without a computer.

My old PSU that burned down was a Fractal Design Ion+ 760W platinum. It's over the recommended 750W power supply specified by the Asus Prime 5070 Ti. When I bought it, I thought I was buying a high-quality, efficient and safe PSU. It's still under warranty for years to come. I had the 70Ti undervolted and slightly underclocked, so it was only using 280W even under stress test. This is under the slightly above 300W it should be using at stock.

I used the 3x 8-pin to 12-pin adapter that came with the card, and I made sure that I hooked it up to three independent power cables that go into three different slots of the PSU. I had to do this, since my old PSU did not have a 12-pin cable natively. I did not daisy chain any of the three. I checked the cables and the adapter for damage after I removed the card, but they all seemed fine. The GPU had a slight scent, though I can't tell if it's burning/melting or just offgassing from the plastic shroud. The PSU still has a strong scent of burnt electronics.

Both theoretically still "work", since I managed to turn the system off before the fire started. I had left the room for about 20 minutes, and when I came back the room reeked of burning electronics and smoke. I have an air purifier and it also was giving me an extreme air pollution warning. PM2.5 concentration of over 444. For reference, the outdoor air quality was ~6 when this happened, and the previosuly measured room air quality was ~1. Clearly something was burning or melting, and if I didn't catch it right then (and I was thinking of leaving the house with FurMark running), my house could have burned down.

I don't know what to blame. I believe the PSU had an issue, as even if the GPU was faulty, the overcurrent, overvoltage and overtemperature protections should have kicked in far before any melting or burning started happening. Perhaps it was just the PSU that was faulty, but since Asus and nVidia agreed the idle power consumption was not normal even before the near-fire, I dare not try it again with my new PSU. Before I installed this one, I also ran FurMark on my old GPU and both the old card and PSU handled it just fine. Granted, only for about 10-15 minutes instead of 20, and that only ran about 200W, as opposed to this one's 280, but still.

I also bought a wall meter. I managed to test the 50Ti before this incident. Comparing it now with my old card installed again, the wall meter appears to confirm that the idle power draw issue was present and measured correctly. 70Ti showed readings of ~20-25W more than current (old) card.

I don't want to fearmonger anyone, but I should use this as a caution to anyone experiencing unusual power or temperature readings to contact customer support and get their hardware sorted. Of course, "unusual" should only be compared with the same exact card, as for example some 70Tis normally idle at ~28W. A few MSI cards even idled at 40, though this appears to now be solved. Also maybe avoid open box GPUs. Oh, and don't leave stress tests unattended, at least for new cards.

I hope my experience can help someone out. If anyone has any questions, I'd be happy to answer.