RTX 5080 - performance fluctuation

[defu24]

I recently got an ASUS TUF 5080, and i have some weird fluctuation in performance in some of the games i`ve tried so far (MSFS 2020/24, Indiana Jones, COD BO6). Everytime the game starts running smoothly, with decent FPS and no stuttering, and after 20-30 minutes of gameplay (sometimes more but other times even less), the fps drops heavily, input lag rises significantly, and the weird part is that even tough the GPU usage is at 100% when this happens, the GPU temps drops from around 60 degrees to around 40. Sometimes it resolves itself after 2-3 minutes, and then runs smoothly again, other times the only solution is to restart the game. So i`ve installed GPUz to see if i can find anything, and first of all i`m not missing any ROPs, and then i tried to watch what happens when the stuttering begins. As you can also see in the GPU-Z sensor log i`ve attached (while in MSFS 2024), the GPU Load stays pretty much at 100% with ~15GB Vram usage, but the board power draw drops from ~300W to only ~100W, and the temps from ~60 to ~40-50 degrees. I`ve also noticed that the memory controller load is around 5% when the stuttering occurs, while is around 20% when it all runs ok (i`m not sure what percentage the memory controller load should be in normal situations).

It happens both in DLSS mode or TAA, with frame generation on or off.

Other specs are: Ryzen 9800x3d, on an ASUS Strix 870E, a Samsung 990 PRO SSD and the PSU is Corsair 1000x RM.

I also have all the drivers up to date, from the GPU, to the latest BIOS, and motherboard drivers.

Any idea or suggestion of what it might causes it?

Thank you!

 

[tugrul_LordOfDrinks]

Check vram usage and ram usage. Check if windows is using disk for storing ram. Check windows updates or antivirus if they are running on background.

 

[agent_x007]

From what I see, you are running out of VRAM (or drive/game decides that you are and uses NVMe more - in witch case a fix is needed in driver/game software not on your side).

Drop quality settings and/or resolution (to always be below 13GB VRAM used), maybe use two notches lower quality DLSS ?

 

[trandeira]

I recently got an ASUS TUF 5080, and i have some weird fluctuation in performance in some of the games i`ve tried so far (MSFS 2020/24, Indiana Jones, COD BO6). Everytime the game starts running smoothly, with decent FPS and no stuttering, and after 20-30 minutes of gameplay (sometimes more but other times even less), the fps drops heavily, input lag rises significantly, and the weird part is that even tough the GPU usage is at 100% when this happens, the GPU temps drops from around 60 degrees to around 40. Sometimes it resolves itself after 2-3 minutes, and then runs smoothly again, other times the only solution is to restart the game. So i`ve installed GPUz to see if i can find anything, and first of all i`m not missing any ROPs, and then i tried to watch what happens when the stuttering begins. As you can also see in the GPU-Z sensor log i`ve attached (while in MSFS 2024), the GPU Load stays pretty much at 100% with ~15GB Vram usage, but the board power draw drops from ~300W to only ~100W, and the temps from ~60 to ~40-50 degrees. I`ve also noticed that the memory controller load is around 5% when the stuttering occurs, while is around 20% when it all runs ok (i`m not sure what percentage the memory controller load should be in normal situations).

It happens both in DLSS mode or TAA, with frame generation on or off.

Other specs are: Ryzen 9800x3d, on an ASUS Strix 870E, a Samsung 990 PRO SSD and the PSU is Corsair 1000x RM.

I also have all the drivers up to date, from the GPU, to the latest BIOS, and motherboard drivers.

Any idea or suggestion of what it might causes it?

Thank you!

hi!! i have the same problem but i dont know why happen... Have you found the solution, or do you know what causes this error?

 

[tugrul_LordOfDrinks]

I have 3 theories

- pcie bridge heats
- the heat changes the resistivity of pins due to expanding metal under heat and some dust
- then pcie forces v1.1 x1 rather than 5.0 x16
- decreased pcie bandwidth lowers gpu load

Or,

- The games or driver leaks memory
- vram fills
- vram spills over pcie at 50GB/s bandwidth which is 1/20 of vram bandwidth

Or,

- you get bored in middle of game
- start watching star trek new generations from a compressed file using gpu acceleration, at 8k upscaled resolution

 

[trandeira]

I have two theories

- pcie bridge heats
- the heat changes the resistivity of pins due to expanding metal under heat and some dust
- then pcie forces v1.1 x1 rather than 5.0 x16
- decreased pcie bandwidth lowers gpu load

Or,

- The games or driver leaks memory
- Rellenos de VRAM
- vram spills over pcie at 50GB/s bandwidth which is 1/20 of vram bandwidth

hi!
Thanks for answering. It only happens to me in two games: FF XVI and Assassin's Creed Shadow.
The other day I played Baldur's Gate 3 for 5 hours straight and didn't experience a single FPS drop or problem. I don't know, it's just weird.
maybe drivers issue...

 

[tugrul_LordOfDrinks]

hi!
Thanks for answering. It only happens to me in two games: FF XVI and Assassin's Creed Shadow.
The other day I played Baldur's Gate 3 for 5 hours straight and didn't experience a single FPS drop or problem. I don't know, it's just weird.
maybe drivers issue...

Also try to test aliens fire team elite with 200% maxed graphics. It works like furmark. It makes card so hot that I cant even touch the backplate longer than 1 second.

 

[trandeira]

Also try to test aliens fire team elite with 200% maxed graphics. It works like furmark. It makes card so hot that I cant even touch the backplate longer than 1 second.

I try this game in 4k and 200% and works well 280 W constantli... black myth wukong the same... so i dont know what can be the issue, driver nvidia?

 

[tugrul_LordOfDrinks]

Maybe hot spot temperature high due to dislocated cooler or pads?

 

[trandeira]

Maybe hot spot temperature high due to dislocated cooler or pads?

nah in hwinfo64 hotspot temp its like 65/70 so this not the problem.
But I've noticed that when that happens, "Performance Limit: Reliability Voltage" pops up in hwinfo64 and the game goes from 60 fps to 45 power from 250W to 145W and temp from 60 to 42
but just in that game... wukong setup maximum not happens so i think its drivers still not good for series 50

Edit:
graphics cards have any fps limiter?
I have seen that it limits to 40/45 fps i change 4k to 1080p and still 40 fps so the problem its fps no the power energy consumption