GPU Voltage fluctuates madly

Technical Support
1

So I finally decided to try diablo 4 again after safely testing the beta with my GPU (at that time I didn’t even know D4 was bricking GPUs) - it was a complete leap of faith.

My setup, configurations and tinfoil hat measures:
GPU: MSI RTX 3080 Ti
CPU: Ryzen 9 5900x
RAM: DDR4 128GB
PSU: 1000W Seasonic Prime Platinum
Mobo: MSI MAG X570 Tomahawk Wifi
Set FPS cap in Nvidia control panel to 60 FPS
Set FPS Foreground limit in-game to 60 FPS
Disable Nvidia Reflex Low Latency in-game AND in Nvidia control panel (Low Latency = OFF) - This setting really helps a ton.
I set my graphics quality preset down to even medium in-game
Undervolted my GPU in MSI after burner
Set power limit in MSI After burner
Resolution is set to 1440p (I’m playing only in a 1440p monitor anyways)
Tinfoil measure: I installed D4 on a HDD instead on a SSD as someone suggested here, supposedly it helps. I do think it does :slight_smile:

When an average user has to do all these hoops just to play your game, you can tell something isn’t right already.
I would dread to fathom to play D4 without all of these settings. My GPU would probably end up in the casualty list, so I refuse to test it without these settings applied.

But before that, I would like to explain why you want the GPU voltage to be as stable as possible. If your GPU voltage fluctuates, it adds wear to the choke of your gpu. It is fine if the fluctuation is only like 0.050V to 0.080V, but if it is a sudden rise and drop by as much as +0.2V, that’s insane. It’s a 26.67% increase in voltage from 0.750V. This is exactly what’s happening in Diablo 4 as I recorded it on MSI afterburner. I think the upper limit of voltage for the RTX 3080 Ti is 1.1V, but anyhow you wouldn’t want to reach that voltage or fluctuate to that voltage so frequently, because it adds wear and tear to your GPU, particularly on the chokes of your GPU.

Imagine this - your wall socket is rated 110V in North America, then it suddenly fluctuates to as much as 140V (applying that same 26.67% increase) and then it drops to 130V, can you comfortably say that your appliances plugged into that socket would still be safe?

Or another analogy for those unfamiliar with electricity. If you have a pipe for water to pass - you pump the normal amount of water it can handle, the pipe doesn’t burst.
But if you suddenly were to pump even more water through it (like 2-3x the amount of water), then you’re adding stress to that pipe. It’s the same thing that’s happening to your GPU as you play D4, because it is poorly optimized. Simply put, there’s something wrong with the game and that can be fixed. I don’t buy the “Your gpu is about to die anyways” crap and “you can never blame games as games don’t break gpus” bs without thoroughly investigating all the possible factors. If the game devs cannot address the source of the problem, which is the crazy amount of work or render that they’re asking everyone’s gpus to do is absurd, then the problem will still exist. It’s akin to diabetes - where someone that eats a ton of sugar gets insulin resistance because the body is sick and tired of so much sugar you’re putting into the system - and then if you throw in the argument: “It’s not the person’s fault for eating more sugary food”, then you’re just ignoring the source of the problem because he can’t control what he puts inside his mouth - which is the source of the problem.

The voltage of my GPU fluctuates very wildly. It ranges from 0.750V to 0.968V (at one point it even jumped from 0.750V to 1V). I was testing it last night but forgot to record it, but when I went to the inventory screen, the voltage on after burner went from 0.768V to 0.968V - a BIG jump. Its not reasonable that your GPU should be put under stress each time you need to open your inventory screen. This game is poorly optimized and is adding a lot of wear to everyone’s GPUs playing this game.

It’s not just in the inventory screen as well, but I notice that every fire you come near to (such as a brazier or camp fire site), the voltage increases wildly, especially the one near Rovna.

This explains to me why pretty much all GPUs are vulnerable and being put to this unnecessary stress of a dumpster fire. It’s not just the Gigabyte RTX 3080 Tis, EVGA 3080 Tis - but I’ve also read other brands and models such as the 3080 Ti Vision, someone’s MSI 4090 having problems, 2070 Supers having problems, etc. are running into crashes and possibly getting bricked in this forum.

I’ve decided to just quit the game until they fix this issue.

2

You do know that the voltage will fluctuate according to the demand on the GPU, right?

Please tell me you don’t think the GPU is supposed to stay at one voltage level???

Both CPU and GPU do this. The thing is, if you are undervolting, you are essentially trying to cap the amount of voltage it can supply at its peak.

Take my MSI laptop that I have been running on all weekend. (on a trip)

CPU voltage:

.690 - 1.250

GPU voltage:

.575 - .844

But its a mobile 3050Ti. 45W card. Hits 190F on the hot spot. Plays the game fairly well on high settings at 60FPS. Being the card has less vram, and a choke of a memory bus, it does stagger occasionally when teleporting to town, but over all, plays pretty good. Has not hit thermal limits, but is constantly at 100% utilized both GPU and Memory.

Despite that, completely playable. But my Desktop back home just plays this game so much smoother and with less heat.

Its an i9 12th gen, 32GB DDR5, 3070Ti. Very high speed NVME as well. Have played on it for several hours at a time, no problems.

I have a single fan design on this laptop. MSI Thin GF63 series. 11th gen i5, 32GB of memory and a 1TB of NVME drive. But quite capable for this game. Played about 3 hours today in fact.

As for cards getting bricked, that’s happening because the cards have an issue. And Gigabyte have been the worst examples. From cheap capacitor designs, to boards that crack and weaken tracings near the PCIe connector, to just general poor quality. Especially in their 3080/3090 line, which is sad. Diablo IV just happens to be stressing them hard enough to trigger the failures.

Anyway.

3

Yes I know but you miss the point.
It fluctuates by a LARGE amount and FREQUENTLY
That’s a problem with this game.

This is the only game I have ever played that fluctuates by as much as +0.2V to even 0.25V that’s very harmful to the chokes of your gpu. Google why large voltage fluctuations is harmful to electronics, you will understand what I mean.

4

I think you are reading more into it than you should. I know it spikes, along with power, utilization, and heat when opening vendor screens, and drop a ton when you open the map.

But the hardware is built to handle this. Now if you have a PSU that is cheap or under-powered or defective, that might be a cause for concern.

5

Then that would mean the devs ignoring the problem as well.
I wouldn’t put so much faith in hardware if its constantly being abused as it shouldn’t be. I’m pretty sure the devs could just tone down whatever they’re asking the gpu to render so that this can fix the problem easily and not cause anymore stress to everyone’s gpus. They surely missed the QA part of testing this game with afterburner.

6

Yeah, I was using Afterburner myself in this video here a few days ago on my desktop:

That also includes some thoughts and suggestions.

7

The electrical engineer in me wants to spend an hour explaining why this post is extremely flawed and outright wrong in some spots, but it’s almost midnight and I’m being lazy… You’re talking about “wear this out” “wear that out” etc etc, but I will leave it with a one word explanation: Tolerances.

Also, you’re talking about chokes… Do you know what a choke is? It’s an inductor… Do you know what an inductor is? A coil of enameled wire… You’re acting like it’s some electrolytic capacitor that breaks down from hot/cold cycles.

The voltages oscillating all over the place is 100% normal and that’s what they are supposed to do to min/max performance vs power consumption. Otherwise, your GPU would sit at 300 watts non-stop, no matter how loaded it actually was. Now THAT will wear out components FAAAAAAAAaaaaar faster.

Stop falling for Youtube clickbait videos and buzzwords made by people that have no clue what they are actually talking about.

3 Likes
8

Yes, we are observing the same phenomenon - as you put it “stuff goes up if you open the inventory or menu” in your video which shouldn’t be the case.
Even if our GPUs are different and our settings are different, we still experienced the same result.

So simply put a problem does exist. We are just using different metrics to observe the same problem. I used voltage, but you used wattage and temperature in your video - but they’re all interrelated in one equation: Power (watts) = Voltage x Current (Amps)

I don’t care how others explain it to me, but it satisfies me that you do belong in the camp that actually acknowledges there is a problem in the game and that the devs should do something about it.

9

Oh yeah, the game has issues. But what you were describing is normal operation of a video card, even if it seems extreme. So while I have seen the same voltage changes, I wasn’t nearly as worried about those as I was about the memory running higher temps than the GPU or hotspot. That actually concerns me more. But at least I have set my system up to keep the card overall at a lower operating temp. Now my laptop on the other hand, I could use for a space heater, but at least it hasn’t been thermally throttling itself yet.

10

See my post about this and why it’s happening:

11

The problem is that these characters are being rendered with what looks like the same full suite of features as the rest of the game, in terms of lighting and shadows, and are also at very high resolution+quality. So basically, the game is rendering the game world scene

Yes, this is what I also stated in my video if you’ve watched it. A game rendering within a game if you open your inventory, with at least the same quality. Unnecessary amount of work. More work needed = GPU’s voltage goes up = Power (watts) go up = Temp goes up. Devs should tone the amount of work down. This is what happens if you launch a game half-baked for a quick cash grab.

12

Nah, they just built the game around 90% of players playing at a vsynced 60fps between the console versions and the bulk of PC gamers. If you lock your PC to 60, it mostly resolves the problem, even on lower end GPUs that don’t have a ton of headroom. On overkill GPUs, you can lock higher, like 90 or 100. Take your raw FPS maxes and cut them down to 2/3 or so. So if you can run at 150fps, lock to 100. This will clamp the hell out of the spiking.

But as I said in my post that I linked, they should just clamp all 3D UI elements to only update at 30 or 60fps.

13

I 100% agree with tou, theres something wrong with the game. Why is the voltage going up if im just opening a vendor? Makes no sense

1 Like
14

Again: Inventory screens increase GPU load—>Increased GPU load needs more power to compute—>The GPU draws more power—>P=VI—>If P has to increase, VI has to increase—>The components like the CPU and GPU typically handle most of their power regulation through the modulation of the voltage.

This is why so many people are into undervolting: You get the same “power” for less power consumption and is also why it can lead into stability issues if you UV too far(qualifying logic voltage levels for a “1” can dip into the indeterminate range, leading to errors).

15

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.