Help Needed: GPU Clock and Power Draw Issues (RTX 3070 Ti Mobile)

[matias.comesana]

My laptop is a Gigabyte Aorus 17 XE4, which I bought in October 2022 (it's not under warranty anymore). It comes with an RTX 3070 Ti. The problem is that very often, the GPU clock gets stuck at 210 MHz, and most importantly, the board power draw increases and gets stuck at 752 W, which is obviously a bad reading or something, as there's no way it is actually consuming that amount of power.

I took 2 screenshots and a full video showing the problem while using FurMark (to stress the GPU), GPU-Z, and HWMonitor at the same time. These 2 screenshots show the "before" and "after" of the exact moment where the clock decreases from 1245 MHz to 210 MHz, while the power increases from 128.6 W to 752 W. When this happens, the laptop remains with these values until I shut it down or reset it. Sometimes, when I start the PC and after using it for a while with non-demanding tasks like Chrome, Outlook, or some other non-gaming activities, I check GPU-Z and find that it’s stuck at 752 W.

So far, I've tried uninstalling GeForce Experience, installing older GPU drivers from the Gigabyte website, and nothing seems to solve the problem. I also have MSI Afterburner installed. I don't recall when the problem started. I've never flashed the vBIOS.

I don't think the problem is the thermal paste since it doesn’t explain the sudden bad readings from GPU-Z. Perhaps it is a faulty sensor? But it's strange that a faulty sensor gets stuck with faulty values; I would expect a faulty sensor to give us faulty readings all the time, but who knows. Really, I'm by no means an expert, so I need your help, guys.

What do you think the problem is? What should I do first? DDU and install GPU drivers from the Nvidia page? Reinstall Windows? Flash the vBIOS (I don’t know where to find the most recent and compatible ROM in such a case)? Change the thermal paste?

Before

After

Full Video

 

[eidairaman1]

Faulty sensors on the gpu/motherboard/cpu is plausible, and may need a hardware repair, did you do a fresh OS?

 

[matias.comesana]

Yes I did and didn't work . I tested with stock Nvidia drivers from Gigabyte's support page and found a strange behavior: if I used furmark or if I plugged the displayport the power draw will show 752w, but if I closed furmark or plugged off the displayport cable it will go back to normal power draw values. Can it be something related to vbios or bad drivers? I think its related to sensors because some months ago I played tons of games (some of them demanding like Spiderman) and I didnt have any issues at all. I will send it to the official Gigabyte repair center in my town (Montevideo, Uruguay) but with little or no hope of the problem being solved...

 

[lexluthermiester]

My laptop is a Gigabyte Aorus 17 XE4, which I bought in October 2022 (it's not under warranty anymore). It comes with an RTX 3070 Ti. The problem is that very often, the GPU clock gets stuck at 210 MHz, and most importantly, the board power draw increases and gets stuck at 752 W, which is obviously a bad reading or something, as there's no way it is actually consuming that amount of power.

I took 2 screenshots and a full video showing the problem while using FurMark (to stress the GPU), GPU-Z, and HWMonitor at the same time. These 2 screenshots show the "before" and "after" of the exact moment where the clock decreases from 1245 MHz to 210 MHz, while the power increases from 128.6 W to 752 W. When this happens, the laptop remains with these values until I shut it down or reset it. Sometimes, when I start the PC and after using it for a while with non-demanding tasks like Chrome, Outlook, or some other non-gaming activities, I check GPU-Z and find that it’s stuck at 752 W.

So far, I've tried uninstalling GeForce Experience, installing older GPU drivers from the Gigabyte website, and nothing seems to solve the problem. I also have MSI Afterburner installed. I don't recall when the problem started. I've never flashed the vBIOS.

I don't think the problem is the thermal paste since it doesn’t explain the sudden bad readings from GPU-Z. Perhaps it is a faulty sensor? But it's strange that a faulty sensor gets stuck with faulty values; I would expect a faulty sensor to give us faulty readings all the time, but who knows. Really, I'm by no means an expert, so I need your help, guys.

What do you think the problem is? What should I do first? DDU and install GPU drivers from the Nvidia page? Reinstall Windows? Flash the vBIOS (I don’t know where to find the most recent and compatible ROM in such a case)? Change the thermal paste?

Before

View attachment 360492

After

View attachment 360493

Full Video

First tip: Quit running Furmark on a laptop. It's causing it to thermal throttle itself. This a known problem.

Second tip: Your system seems fine, see below. If you're not happy with the temps/power draw, try underclocking/undervolting the GPU. You will not feel a difference in performance but you will lower power draw and temps produced by the GPU.

Third tip: Put your laptop up on a stand made for laptops or prop it up by placing a spacer/shim under the back edge of the laptop to give it more room to breath, IE, to improve airflow.

Fourth tip: If you haven't done this already, get a can of compressed air, take the bottom off the laptop(yes, open it up), and completely blow it out.

Fifth tip: Find a way to disable CoPilot. There have been reports of Copilot messing with Windows and Drivers settings without user input. At the moment, it's unconfirmed, but it's not worth risking.