Most games push GPU usage to 100% and crash.

[mRecords]

First of all, english is not my native language, so I apologize if I explain something innaccurately.

I bought a new GPU (specs at the end) and now most games crash somewhere between a couple seconds and a few minutes. They either freeze or just push me back to desktop.
What I did so far:

- I checked the GPU is plugged into the correct socket (by my tech guy). It is.
- I checked temperature with two games opened simultaneously at ultra settings (before they crash). Temp is no higher than 70°C.
- I checked drivers. All up to date. I tried older drivers. Nothing changed.
- I checked BIOS was up to date. It wasn't. Now it is. Nothing changed.
- I checked the event viewer (I hope I translated that correctly) from Windows. It doesn't register any entry.
- Checked that power is enough to feed the GPU. It should be.
- I had W11. Did a full clean up and started from scratch with W10. Still the same.
- Config the GPU from the NVIDIA panel to max performance. Nothing.
- Set the games to the lowest possible settings. They still push the GPU to 100% even on the menu.
- Tried underclocking.

Now what's odd is that it happens with MOST games, but not all. State of Decay 2, DOOM Eternal, both Ori games, Back 4 Blood (and I can keep on going) crash but it doesn't happen with League of Legends (although I suspect that has to do with the fact that it has very low sytem reqs) and, here's the weird thing for me, Grounded to its max settings which stays between 14-27% GPU usage.

Any help would be appreciated.

Specs (everything but the GPU has less than a year old):

Thermaltake Smart 500W power supply.
120GB SSD + 1TB hard drive.
2x 8GB RAM
Gigabyte B450M DS3H V2 Motherboard.
AMD Ryzen 5 5600 processor
NVIDIA gtx 1660 super ventus GPU.

[_I_]

the psu was mfg during the bad cap era
its lucky to have made it this long

running the lower res might not have stressed the 6600 as much as the 1660s
different gpu, different games = different load on psu

[Pepe]

To really eliminate the PSU, you should borrow someone else's working PSU to make some tests... Like I've said, "turbo boost" could induce power spikes greater than normal load.

There can be other reasons, of course. My second guess would be a faulty brand new GPU, it could happen. Test the GPU on another system, test your system with another GPU... To really test a GPU, run FurMark or any other kind of software that pushes your video card to its limits. Possible issues with your video card: GPU chip being broken (v unlikely), one memory module broken, a resistor being faulty, VRM faulty, a capacitor being faulty (most likely).

[Set-115689]

Underclock the gpu is msi afterburner. Memory clock, core clock and power limit. A slight underclock might make the card stable.

[mRecords]

Originally posted by Bad 💀 Motha:

So just so we're all on the same page here.

Can you post an update as to your current issues and what all you've tried.

Nothing to add. I did nothing besides what I already posted and the issues remain the same.

Originally posted by Pepe:

To really eliminate the PSU, you should borrow someone else's working PSU to make some tests.

Yes, I'm working on that to eliminate that factor. I already tried stressing out the GPU with Furmark. I detected no problem.

Originally posted by Set-115689:

Underclock the gpu is msi afterburner. Memory clock, core clock and power limit. A slight underclock might make the card stable.

Already tried. Although I may have not done that right (I'm sort of a cave man). A quick guide on that would be great if anyone can.

------------------------------

I read a comment searching on the web from another couple guys that had the exact same issue but my english is not good enough to catch up with that specific language. Apparently they had a problem with the RAM cards. As far as I understood, one of them had it uncorrectly plugged on the socket. The second guy read that and discovered he had the same problem. Could that be a thing?

[Illusion of Progress]

Furmark's strength is also its weakness. It's great at pushing a GPU to high use, but it's typically a constant use. Games are often more variable/mixed. Furmark is good to include, but I wouldn't start and end testing there.

OCCT has a "GPU variable" test. When I was having issues with my GPU, that was the test that tripped it most (and it didn't even always). If that failed, Furmark (or a game) along with a hardware accelerated browser that I was actively in would also usually trip it before long.

Just adding those as scenarios to possible try.

Originally posted by mRecords:

I read a comment searching on the web from another couple guys that had the exact same issue but my english is not good enough to catch up with that specific language. Apparently they had a problem with the RAM cards. As far as I understood, one of them had it uncorrectly plugged on the socket. The second guy read that and discovered he had the same problem. Could that be a thing?

Reseat your RAM. Is it possible it got unsettled when changing the video card?

I'd also run tests on it with MemTest86.

[mRecords]

Originally posted by Illusion of Progress:

Furmark's strength is also its weakness. It's great at pushing a GPU to high use, but it's typically a constant use. Games are often more variable/mixed. Furmark is good to include, but I wouldn't start and end testing there.

OCCT has a "GPU variable" test. When I was having issues with my GPU, that was the test that tripped it most (and it didn't even always). If that failed, Furmark (or a game) along with a hardware accelerated browser that I was actively in would also usually trip it before long.

Just adding those as scenarios to possible try.

Originally posted by mRecords:

I read a comment searching on the web from another couple guys that had the exact same issue but my english is not good enough to catch up with that specific language. Apparently they had a problem with the RAM cards. As far as I understood, one of them had it uncorrectly plugged on the socket. The second guy read that and discovered he had the same problem. Could that be a thing?

Reseat your RAM. Is it possible it got unsettled when changing the video card?

I'd also run tests on it with MemTest86.

Great, I'll look into that.

[Set-115689]

Install msi after burner (may as well let it install riva tuner also) then on the main screen decrease the core clock by maybe -500 , then the memory clock by maybe -500 and set the power limit to 80(?). Then click on the apply button or it won't get applied. Then play a game. The under clocks can be reduced but may as well try the more extreme under clock first.

Can technically under volt the gpu in msi afterburner.

Does playing less stressful games with the fps capped crash the system?

[mRecords]

Originally posted by Illusion of Progress:

OCCT has a "GPU variable" test.

Reseat your RAM. Is it possible it got unsettled when changing the video card?

I'd also run tests on it with MemTest86.

OCCT showed no errors. I checked both RAMs and they don't seem badly plugged although I'm no expert here. Couldn't make MemTest86 run. I don't understand how to do it.

Originally posted by Set-115689:

Install msi after burner (may as well let it install riva tuner also) then on the main screen decrease the core clock by maybe -500 , then the memory clock by maybe -500 and set the power limit to 80(?).

Does playing less stressful games with the fps capped crash the system?

Did that. No good. The only game that I can play with no problem is League of Legends at 300+ fps. Everything else crashes.I'm not sure what a "less stressful game" could be given that, for example, SOD 2 with the lowest possible set up is basically minecraft.

The main point here is that GPU goes to 100% even on the menu, before it starts loading stuff and rendering graphics inside the actual game. This is true, again, for everything, except Grounded. For some strange reason I can play that on ultra for like half an hour (GPU between 30 and 40%) before it suddenly spikes and goes off for no apparent reason that I can discern.

[Boblin the Goblin]

Originally posted by kitty:

Originally posted by mRecords:

The main point here is that GPU goes to 100% even on the menu, before it starts loading stuff and rendering graphics inside the actual game. This is true, again, for everything, except Grounded. For some strange reason I can play that on ultra for like half an hour (GPU between 30 and 40%) before it suddenly spikes and goes off for no apparent reason that I can discern.

I think we finally figured out what's wrong here: You're playing with vsync off and/or no frame rate limiter. In that situation this 100% GPU usage is expected and that most likely is the cause of your crashing combined with your lower quality power supply.

This is probably your entire issue. Make sure you turn on Vsync in game settings and that should make all of your problems go away.

Vsync won't cause 100% GPU usage Megahurtz.

[Boblin the Goblin]

Originally posted by kitty:

Originally posted by SlowMango:

Vsync won't cause 100% GPU usage.

Playing with vsync off most certainly will cause the video card to be at 100% usage always in every game. That is literally the design function of vsync: To limit the FPS to the monitor's refresh rate and AVOID the video card being at 100% usage always.

It won't do that Megahurtz. I know because I manually turn Vsync off in every game.

[_I_]

Originally posted by SlowMango:

Originally posted by kitty:

I think we finally figured out what's wrong here: You're playing with vsync off and/or no frame rate limiter. In that situation this 100% GPU usage is expected and that most likely is the cause of your crashing combined with your lower quality power supply.

This is probably your entire issue. Make sure you turn on Vsync in game settings and that should make all of your problems go away.

Vsync won't cause 100% GPU usage Megahurtz.

only if its not bottlenecked by the cpu
fps higher than refresh rate will have multiple tear lines, more than one frame composing of a displayed frame

new vsync, will force the gpu to idle after a frame is completed before sending it to the display
vsync off, sends the latest frame data to the display as they are drawn

[Boblin the Goblin]

Originally posted by _I_:

Originally posted by SlowMango:

Vsync won't cause 100% GPU usage Megahurtz.

only if its not bottlenecked by the cpu
fps higher than refresh rate will have multiple tear lines, more than one frame composing of a displayed frame

new vsync, will force the gpu to idle after a frame is completed before sending it to the display
vsync off, sends the latest frame data to the display as they are drawn

It will cause tearing.

[_I_]

yes, thats what i said, multiple tear lines

[_I_]

if the game or driver use old vsync, it will keep drawing instead of going to idle
it may complete 2 frames between refreshes and just display the latest completed one, dropping previous frame

[Boblin the Goblin]

Originally posted by kitty:

Originally posted by SlowMango:

It won't do that Megahurtz. I know because I manually turn Vsync off in every game.

I can do the same thing. I can go play games and turn off vsync and see my video card instantly jump to 100% usage and stay there constantly then turn on vsync and usage drops back to 30% again. I'm not blind. I know what I see. I can see it with my own eyes. Vsync would definitely solve this person's problem. They just need to go use it.

Except it stays level Megahurtz.