Poor Performance in Just One Game

Ask Fedora
, , , , , , ,
1

Alright, let me start off by detailing the rigs…

Machine #1

Gigabyte Z490 Aorus Elite
Intel i7-10700F
EVGA NVIDIA 3060 XC
32gb of DDR4
2TB Western Digital Blue NVMe SSD

Machine #2

Gigabyte Z790 Aorus Elite AX
Intel i9-14900K
ASUS TUF NVIDIA RTX 3080 OC
32gb of DDR5
2TB Western Digital Black NVMe SSD

These machines are both running bone stock Fedora 44 Workstation and the appropriate Nvidia drivers installed through the GNOME software installer from the non-free rpm fusion Nvidia repo. Both machines work flawlessly. They are both lightning fast, ultra responsive and absolutely demolish every single task I can throw at either of them.

A few days ago, a friend of mine asked me to play The Division 2 with them. My 3080 machine was tied up, so I installed steam (not the Flatpak version) onto the 3060 machine and it ran the game pegged at a VSYNC’d 144fps. I did zero configuration, just opened the game and it was smooth sailing.

A few days later, they asked to play again. Since my 3080 machine was free, I installed it the exact same way, launched it the exact same way and ended up presented with a jittery and unstable 60-80fps. Keep in mind, that this rig ran something like 150-180fps in this same game in Windows.

Things I’ve tried..

Enabled/disabled VSYNC in the game
Forced VSYNC on and off and different refresh rates via Steam launch commands
Tested different cables (both DisplayPort and HDMI)
Fiddled with ManguHud and Gamescope
Enabled and disabled GSYNC/FreeSync both in game and in the monitor controls
Disabled efficiency cores.
Tested other monitors
Changed Intel turbo boost from Auto to Forced on
Uninstalled and reinstalled the NVIDIA display drivers.
Tested lots of other games and for instance in Doom Eternal, I saw on average of 230fps without VSYNC.
Wiped the entire machine and reinstalled a fresh copy of Fedora 44 Workstation.

One observation that I did make was that when I was running the game on my 3060 rig, I was basically maxing out both the CPU and GPU. When I run the game on the 3080 rig, I basically see about 20% usage on the CPU and 30% usage on the GPU. I believe it’s an issue with the 14900K staying in a low power state, because I’ve tested the 3080 and it hits ~340 watts and ramps up to the highest power level every time.

I’ve only found one single difference between my 3060 machine and my 3080 machine. When I go into the GNOME power settings and set the power level to “Power Saver”, I see a message that reads “Performance mode temporarily disabled due to high operating temperature”. Keep in mind, I have exceptional cooling and my temps move between 30c and 70c depending on load. Ive also read that this is a known bug that does not affect performance.

I know that was VERY long winded, but I wanted to be clear that I’ve done lots of testing. So, with that being said… Any thoughts?

2

Have you tried different Proton versions?
I use Lutris, and it will automatically update Proton, which sometimes breaks games - so I have to manually set an older or alternative Proton / Wine version.

3

Thanks for input!

Unfortunately, I have. I tried all the stock versions of Proton that ship with Steam, including Hotfix and Experimental. I even tried ProtonGE too…

None of them produced improved results.

4

I would be checking these items.

Did you compare software config between the 2 systems?
Is it the same proton versuon on both systems?
Did you install ganescope type tool on either?
When the gane is running poorly is there any app taking a lot of background cpu time?
Are there any errors in the user or system journals?

5

Thanks for the input!

Yes, I have compared the software config. I even went as far as erasing both computers, installing a fresh copy of Fedora 44 Workstation, and performing the same installs and configuration on both computers.

I also tried Gamescope with no positive results and confirmed that the background CPU time during gameplay is, for all intents and purposes, zero on both computers.

system-compare
system-compare670×862 89 KB

nvidia-compare
nvidia-compare759×688 170 KB

6

Is the power profille the same on both? Performance vs. Low-power etc.

7

Yes, both computers are set to the Performance power mode both in the GNOME settings and in the Nvidia driver settings app (which I don’t think really has any effect when running GNOME 50/Wayland)

8

I would compare CPU core utilization in both scenarios, along with the clock speeds on the cores. If the CPU is throttling itself, it could be bottlenecking the GPU, hence causing the stutters and such. And make sure to update the BIOS if not already updated.

9

That would suggest a problem with CPU cooling if that is happening.

10

Typically yes, but there could also be a bug or some other unexpected issue causing it. Best to measure the underlying performance metrics that you can, when investigating something like this.

11

Yeah, my gut tells me there is some bug in the CPU power management system, because my 14900K is barely utilized in the Division 2. It’s known to be a very CPU intensive game and during a full on busy scene in the game, my CPU is hovering around 20-30% as is my GPU. My GPU is constantly waiting on the CPU to feed it things to calculate.

CPU runs cool, it hovers between 30-70c under normal use and max load respectively. It shouldn’t be anywhere near throttling. And again, it ran at 2.5x the FPS just the other day in Windows.

I think I’m just going to have to live with this performance for the time being and see if updates make any difference over time.

12

Yes, but just note that you should look at it per-core, not overall, to really see what’s going on. The first thing I’d check is how many cores are loaded, and what clocks the cores are reaching. On your CPU, depending on OC/power profiles set in the firmware, the cores handling the game should be hitting 5.5 Ghz+ consistently. You can use whatever monitoring tool you like (e.g. MangoHud with core_load) or just something like watch "grep 'cpu MHz' /proc/cpuinfo" in your terminal.

13

Only in bursts of activity. The sustainable speed is 3.4GHz according to intel https://www.intel.com/content/www/us/en/products/sku/236773/intel-core-i9-processor-14900k-36m-cache-up-to-6-00-ghz/specifications.html
Once the core is hot it will not turbo.

14

Yep, I checked it per core. Both with all 32 cores enabled and with only the 16 p-cores enabled (as sometimes e-cores can cause stutter in certain games). In neither situation, did the system spin up the cores to where they should be.

15

This is inaccurate; the 14900K can sustain 5.5 Ghz+ as long as adequate cooling is provided. Unfortunately Intel’s spec sheet can be confusing with all the marketing terms for their boosting technologies. If interested, you can check out this review for example: Intel Core i9-14900K Review - Reaching for the Performance Crown - Overclocking, Boost & Clock Frequencies | TechPowerUp

What clock speeds were you getting for that game facing the issues? What frequencies do you usually get for gaming workloads?

And did you update the BIOS? It’s especially important for this CPU model, as it is known to get damaged without certain firmware fixes.

Otherwise, I would look into CPU governors, but unfortunately I lack the experience here so can’t advise on specifics. Maybe someone else has ideas.

16

Here is a screenshot of MangoHud. As you can see, the CPU 's just not being tasked with that much of a workload. This screenshot was taken after loading into the game and letting it settle for about 5 minute. There’s a ton of computational potential… The CPU is just not being stressed at all.

To add to this, I erased everything last night, installed Windows 11 Pro and it ran at 160-180 fps. I installed Fedora back onto it and I get this stuttery, jittery mess. I guess I’m just out of luck in this one game.

No biggie. I guess it’s just a Fedora or driver bug.

division2
division21920×804 206 KB

17

Yes, the BIOS is updated to the latest release F0 from Gigabyte, of course. As of June 2026.

bios
bios1650×687 83.4 KB

18

What does Steam support say? I agree it is some kind of Fedora issue - we have seen it before.

19

Yeah, the CPU/software configuration is definitely bottlenecking the GPU somehow. But at least the CPU is hitting the expected 5.7 GHz so it’s not that.
MangoHud with core_load and core_type options would be interesting to see, since it would tell you the load distribution across the different types of cores.
Otherwise, you may need to test different runtimes, e.g. wine vs proton vs proton GE, etc.

20

After digging and digging and digging… I finally figured out what it was… It was the CPU being throttled by a safety/stability measure…

Disabling split lock mitigation got my frame rates up from the mid 70’s to the 180’s…
sudo sysctl -w kernel.split_lock_mitigate=0

I hope this helps someone else that might be having some performance issues like this…

CAUTION: Run this command only once you’ve read and understood it’s implications.