Thanks. I just tried adjusting a few gamescope settings but the utilization still goes down over time. Seems to give me a virtual sandboxed display? I’m honestly not sure how I can use it to troubleshoot more than I can by adjusting the settings in game. Any recommendations on options that might reveal more information?
Thanks again.
I updated my post with graphs from a DLSS Performance + Low Graphics Preset run. Didn’t seem to have any impact, beyond lower initial GPU Load (low 70’s vs high 90s).
Tried as few things from your suggestion:
I might try to run a few other games to see if anything else triggers it. I haven’t noticed any other games with the FPS drop over time but I don’t play many other demanding games.
With my motherboard’s “Expo I” setting (uses timing from RAM rather than system default), I do get occasional core errors in Windows and Linux on hour combined runs, and sometimes just with the CPU/Memory hour run, though its inconsistent - sometimes just one or a few errors, another time a slew of them. I figured my “Expo I” setting might have something to do with it so I turned it off.
With it off I ran it for a full hour, once in Windows and once in Linux, and now receive no errors. Played for about 2 hours tonight on Linux and the issue still occurs without using the “Expo I” setting.
Got about ~110 minutes on Windows 10 tonight. FPS was stable. Graphs posted.
I actually originally started SM2 on Windows 10 last fall playing single player, and I did play a few nights this summer on Windows 10 while troubleshooting to test. I was unable to reproduce the issue on Windows - played 2+ hours without it happening.
I think I did it a few weeks ago actually, when my discord client wouldn’t connect and I figured I’d test it again to see.
If I have time with all my other testing I’ll give it another shot tonight or tomorrow night.
Definitely on Bazzite. I tried lowering the DLSS render scale to performance and also dropping the graphics quality preset.
I’ll double check that it still happens on Tumbleweed though and log it later today.
I don’t click Google Drive links.
Fair enough. I’ve since posted images of several graphs.
I initially assumed thermal throttling or resource exhaustion. I just can’t find any evidence to support either cause.
Heat: No temperatures I can measure are anywhere near the limits that cause throttling. GPU, CPU, SSD are all relatively cool. I can’t measure VRAM temp (because NVIDIA) but it’s also important to note zero issues in Windows.
Resources: Neither RAM or VRAM appear to be exhausted. My idle tests show 5.5/16GB VRAM and 16/64GB System RAM used. Other system-wide monitors confirm that VRAM/RAM and GPU/CPU usage (overall) matches the games usage so nothing external seems to be hogging the resources. I have plenty of space left in my partitions so I’m not running out of disk space.
I ran the game for a while, idling. GPU load is down to 70%. I ran the process. Also tried looping nvidia-smi in the background with the game as the focus for a while.
> nvidia-smi -q -l 1 -f smi.log -d PERFORMANCE
All of the records look like this:
> nvidia-smi -q -d PERFORMANCE
==============NVSMI LOG==============
Timestamp : Tue Sep 23 07:06:07 2025
Driver Version : 580.82.07
CUDA Version : 13.0
Attached GPUs : 1
GPU 00000000:01:00.0
Performance State : P0
Clocks Event Reasons
Idle : Not Active
Applications Clocks Setting : Not Active
SW Power Cap : Not Active
HW Slowdown : Not Active
HW Thermal Slowdown : Not Active
HW Power Brake Slowdown : Not Active
Sync Boost : Not Active
SW Thermal Slowdown : Not Active
Display Clock Setting : Not Active
Clocks Event Reasons Counters
SW Power Capping : 1160456 us
Sync Boost : 0 us
SW Thermal Slowdown : 0 us
HW Thermal Slowdown : 0 us
HW Power Braking : 0 us
Sparse Operation Mode : N/A
Since I’ve been checking, my kernel has been 6.14 to 6.16+, whatever Bazzite and now openSUSE Tumbleweed defaults to. Currently - Kernel Version: 6.16.7-1-default (64-bit). I’m not familiar with the xanmod kernel or running custom kernels at all.
I believe that I’ve ruled that out. See the blue line in the fourth fifth graph in each set. When idling VRAM is steady at ~5.5GB, but performance degrades. I’ve also checked VRAM availability outside of mangohud while the game is running and my total VRAM usage/availability aligns with the game’s VRAM usage reported by mangohud.
It seems I get stuck once a benchmark is finished. I downloaded what appears to be an appimage from their website. Ex.
00:00:00 - Info - Benchmark schedule started at 2025-09-22 23:46:03
00:00:00 - Info - CPU - Benchmark initializing
00:00:00 - Info - CPU - Benchmark started
00:00:10 - Info - CPU - 1 threads, SSE : 108.47
00:00:10 - Info - CPU - Benchmark stopped
00:00:10 - Info - CPU - Benchmark initializing
00:00:10 - Info - CPU - Benchmark started
00:00:20 - Info - CPU - 16 threads, SSE : 954.19
00:00:20 - Info - CPU - Benchmark stopped
00:00:21 - Info - CPU - Benchmark initializing
00:00:21 - Info - CPU - Benchmark started
00:00:31 - Info - CPU - 1 threads, AVX : 229.73
00:00:31 - Info - CPU - Benchmark stopped
00:00:31 - Info - CPU - Benchmark initializing
00:00:31 - Info - CPU - Benchmark started
00:00:42 - Info - CPU - 16 threads, AVX : 1 935.24
00:00:42 - Info - CPU - Benchmark stopped
00:00:42 - Info - Uploading the results to Ocbase
00:00:42 - Info - Schedule Completed
~~~~~
00:00:00 - Info - Benchmark schedule started at 2025-09-22 23:52:29
00:00:00 - Info - Memory - Benchmark initializing
00:00:00 - Info - Memory - Benchmark started
00:00:10 - Info - Memory - Read : 2 034.06
00:00:10 - Info - Memory - Benchmark stopped
00:00:10 - Info - Memory - Benchmark initializing
00:00:10 - Info - Memory - Benchmark started
00:00:20 - Info - Memory - Write : 780.88
00:00:20 - Info - Memory - Benchmark stopped
00:00:21 - Info - Memory - Benchmark initializing
00:00:21 - Info - Memory - Benchmark started
00:00:31 - Info - Memory - Combined : 1 317.76
00:00:31 - Info - Memory - Benchmark stopped
00:00:31 - Info - Uploading the results to Ocbase
00:00:31 - Info - Schedule Completed
~~~~~
00:00:00 - Info - Benchmark schedule started at 2025-09-22 23:56:51
00:00:00 - Info - Storage benchmark - Benchmark initializing
00:00:00 - Info - Storage benchmark - Benchmark started
00:00:21 - Info - Storage benchmark - Sequential Read : 7 360.58 MB/s
00:00:21 - Info - Storage benchmark - Benchmark stopped
00:00:22 - Info - Storage benchmark - Benchmark initializing
00:00:22 - Info - Storage benchmark - Benchmark started
00:00:43 - Info - Storage benchmark - Sequential Write : 6 246.57 MB/s
00:00:43 - Info - Storage benchmark - Benchmark stopped
00:00:44 - Info - Storage benchmark - Benchmark initializing
00:00:44 - Info - Storage benchmark - Benchmark started
00:01:06 - Info - Storage benchmark - Random Read : 1 651.07 MB/s
00:01:06 - Info - Storage benchmark - Benchmark stopped
00:01:07 - Info - Storage benchmark - Benchmark initializing
00:01:07 - Info - Storage benchmark - Benchmark started
00:01:28 - Info - Storage benchmark - Random Write : 1 870.32 MB/s
00:01:28 - Info - Storage benchmark - Benchmark stopped
00:01:29 - Info - Schedule Completed
00:01:29 - Info - Uploading the results to Ocbase
After the CPU, Memory, and Storage benchmarks it consistently just… sits. No errors, nothing of note. I have a big stop button and the green animation continues at the top left. However nothing else happens after this point if I wait/leave it for a while. The “Stop” button is unresponsive. I have to close and reopen. Am I doing something wrong? I tried Memory and CPU multiple times, same behavior. Tried the “Test” version 15, same.
I did have some success with the latency test. Only tried it once though.
00:00:00 - Info - Benchmark schedule started at 2025-09-22 23:54:02
00:00:00 - Info - Latency / bandwidth - Benchmark initializing
00:00:00 - Info - Latency / bandwidth - Pausing monitoring to ensure accurate readings through the benchmark
00:00:00 - Info - Latency / bandwidth - Benchmark started
00:00:03 - Info - Latency / bandwidth - L1 Latency : 2.89 ns
00:00:10 - Info - Latency / bandwidth - L1 Read Bandwidth : 2 557.30 GiB/s
00:00:17 - Info - Latency / bandwidth - L1 Write Bandwidth : 2 550.08 GiB/s
00:00:20 - Info - Latency / bandwidth - L2 Latency : 6.72 ns
00:00:27 - Info - Latency / bandwidth - L2 Read Bandwidth : 1 785.33 GiB/s
00:00:34 - Info - Latency / bandwidth - L2 Write Bandwidth : 1 745.18 GiB/s
00:00:40 - Info - Latency / bandwidth - L3 Latency : 51.07 ns
00:00:46 - Info - Latency / bandwidth - L3 Read Bandwidth : 593.72 GiB/s
00:00:53 - Info - Latency / bandwidth - L3 Write Bandwidth : 356.05 GiB/s
00:01:01 - Info - Latency / bandwidth - Memory Latency : 73.82 ns
00:01:13 - Info - Latency / bandwidth - Memory Read Bandwidth : 82.86 GiB/s
00:01:24 - Info - Latency / bandwidth - Memory Write Bandwidth : 30.44 GiB/s
00:01:24 - Info - Latency / bandwidth - Benchmark stopped
00:01:24 - Info - Latency / bandwidth - Resuming Monitoring
00:01:24 - Info - Uploading the results to Ocbase
00:01:24 - Info - Schedule Completed
00:01:24 - Info - Upload successful
Running the stability now and going to sleep.
I’m in IT and I was a computer Tech. I enjoy building gaming PCs and I know a good bit but this is really frustrating me. As I said, I’m currently driven by spite and hatred of Windows, especially 11.
I’ll play a bit and keep an eye on my SSD temp and I’ll run OCCT later. Thanks!
Mission Control? Do you mean Mission Center?
Sorry, wasn’t sure which SSD I had the Windows Partition on and needed to look up commands to check. Looks like I’m using the newer Sandisk WD Black SN850X for Linux.
Yes. I can’t find anything beyond my own post complaining of my specific issue. I’ve also added more hardware detail and images of a few of my graphs since my link seemed to be ignored.
Cooler, storage and OS?
Added more detail to the Hardware Spoiler. NOCTUA NH-D15S. Temps don’t seem to be an issue - as I mentioned in the post, usage/load goes down over time, eventually so much that so do my temps.
As mentioned in the post, OS was Bazzite, trying Tumbleweed now on a whim to see if it was a Bazzite issue. It’s not, still happans.
First, always check ProtonDB and see if other people have the same issues or tweaks to fix.
I’ve edited my post for clarity. I’ve checked for similar reports and posted my own experience on Proton DB. I described what I found in the beginning of my post - nothing specifying decreased performance over time, at least the last time I checked.
Second, get some system metrics, and check logs to see if anything obvious is showing.
Did… you look at the graphs? There are a lot of them. I’ll see if I can post screenshots here - I’ve not tried that before.
Edit: Are there some system logs I can check for more information?
While Steam is more or less the best big solution we have, it does leave a lot to be desired. The only reason they are the best is because they clawed their way to the top early, kept themselves “good enough” compared to the competition, and haven’t yet sold out their entire customer base.
At this point, they completely dominate. It’s insanely difficult to compete with them. So long as they make half of an effort to improve things and continue to be somewhat benevolent they’ll likely keep their crown.
However, Valve is not ideal. They are still looking out for themselves, primarily. Many of Valves improvements have just been reactions to competitors and other threats not an inherent desire to deliver the best product possible or do the right thing. It’s just the fact that most competitors are more obviously greedy and immoral that makes Valve look like the heroes.
Without Epic and others throwing cash on the fire trying to compete I doubt we’d have seen even the slow upgrades to the Steam experience we’ve seen in recent years.
Without the Australian lawsuit, we’d have no return policy.
Without the clever abuse of arbitration by a group of lawyers, Valve would still have forced arbitration in the agreements.
Steam OS was only a thing, and Proton only got backed by Valve, when Microsoft first started positioning itself to eat Valve’s lunch by exerting control over Windows and pushing for things like UWP and the MS/Windows/XBOX storefronts on PC.
The vast majority of Valve’s storefront improvements are algorithms and crowd sourcing solutions. They want to be as hands off as possible because being hands on is hard and comes with liability. The whole skins market and gambling fiasco kind of shows that they’d much rather just not get involved if possible - same risk/reward cost/benefit analysis used by every greedy company. If that means lying about how aware they are of it that’s what they’ll do.
Don’t get me wrong. The least worst is, unfortunately, the best we’ve got. I love gaming and use Steam a lot. It’s just that the other big players are just so terrible that I think Valve gets a free pass. Hell, much of the tech industry is swallowing tactical nukes hoping that the radiation will somehow mutate them into a good business. In the meantime they are using the illusion of “expansion” from the resulting explosions to make themselves look bigger for investors. I support anyone not doing that.