Major performance (VRAM leak likely) issue triggered mid-matches in recent updates

[Jill]

Some time after a recent update (unsure which one) I started having major performance issues. There seems to be something specific triggering this, as the game runs fine until mid-match at which point the FPS suddenly plummets (I believe this may be accompanied by a significant spike in memory usage, but I'm not sure.) I can't place exactly when this began happening, but the game was fine until some time this month to my recollection.

During this time, the FPS counter will also sometimes end up bugged and be stuck at a specific number which is not reflective of the actual FPS.

This FPS drop will occasionally clear up after a short amount of time but in most cases it will persist until the game is restarted or the match ends.

I've yet to find a proper common denominator between the instances of it occurring, but it's been happening in every match I've played today and necessitated a restart of the game to correct it.

OS: Windows
GPU: GeForce RTX 2070 SUPER
CPU: AMD Ryzen 5 3600
NVIDA driver version: 576.88 (I have not tried updating, I will report back after I have had the chance)
Renderer: Vulkan
Settings:


I did play one match after switching from Vulkan to D3D11 and did not experience the issue, but I may have just gotten lucky and not encountered whatever causes it. I will test more matches with D3D11 when I am able. I will then update my drivers and re-test on both renderers.

Annoyingly, this problem does appear to be system-dependent. My friends have not been having any issues at all, but one of them does report that one of their friends also has this issue.

I've also not had a chance to experiment with different settings - I recall one bug some time ago where having distance field shadows enable would cause major performance dips for a prolonged time if Vyper died while sliding. I have no idea if it could be a similar problem but I will also test with various different settings when possible

I've not been able to reproduce the issue in replays with any configuration thus far, so I don't have exact timestamps for most of the matches where it occurred (in the future, I will take a note of the match timestamp in which the lag begins/ends and post more examples) but the issue did occur in the following matches:
46808032 (Happened right at the end of the match as my team entered the patron pit, around 47:50 in the replay)
46810901
46813087
46817365
46820196

There are a LOT of similarities to this issue, so I'm going to monitor VRAM usage to see if the same symptoms occur. There are some key differences though in that the users in this thread are on Linux whereas I am on Windows. Furthermore, this was happening as far back as August this year for these users but for me I've never had the issue until very recently. It seems as if the issue has somehow been fixed on Linux but transferred to Windows? Assuming it's the same problem. It would be quite eerie if this issue began manifesting on Windows around when it was fixed on Linux (the timeframes may genuinely match up here as far as I can see)

 

[Jill]

As an update on this I've tested a couple things.

I didn't experience the same issue on DX11. However I did notice that after a point the game began hitching seemingly at random. I have no idea if this is related - it's possible the same issue is causing both and it's just manifesting differently.

I've also turned off distance field shadows but noticed no change, so it's unlikely to be related.

I was having a fairly smooth session after switching back to Vulkan but the issue did eventually rear its ugly head again. I wasn't able to check the memory statistics because turning on cl_showmem caused me to crash.

 

[Jill]

Happened throughout the match in 47019118. Noticeably spiked when our patron got dropped (though was happening on and off before that) and got worse when I looked towards the middle of the map. Mitigated by looking straight down, or elsewhere.

Got REALLY bad to the point that the game was barely playable 3-5 minutes or so before the end of the match.

Did seem to generally correspond to high intensity moments though sometimes would just be kind of random. At the end of the match it was just kind of constant.

I've noticed that when it starts happening, UI textures begin struggling to load in and often show up super blurry before eventually showing up.

I honestly can't decide whether I prefer the frame tanks from Vulkan or the hitches from D3D11.

 

[Jill]

47054477 was very interesting.

Started lagging around 24:04 but seemed to end as soon as McGinnis' med specter particle appeared. No idea if there was a correlation.

Started again and persisted until the end of the match as soon as the teammate I was spectating walked into the enemy base right near the end.

Both instances happened around the enemy's base entrance but I don't know if there's a correlation here or if it's just because those tend to be high intensity spots on the map. I'd note that the rest of the match was completely fine. It's not likely to be super related because I've had this issue crop up as early as laning phase.

cl_showmem still crashes for me but GPU-Z did show that my GPU memory usage was almost maxxed out which seems to fit the idea that this is a video memory leak.

 

[pauljuff]

You can press F7 to show a console, and see if any logs popup exactly as the lag is happening and post here maybe with a video. thats how I fixed my stutter, it was a mic that keeps updating

 

[14NKOS]

This issue happens to me to. I first noticed it back in 2024. Back then i've been experiencing same bug on DX11 but after some update DX11 became stable. As far as i've noticed game starts using significantly less RAM mid match (notice that this info is outdated and last time i checked was back in april 2025). It's really frustrating because Vulkan generally gives me smoother gameplay with more fps till about 22 minute mark. So i switched to DX11 for now.

 

[Jill]

As an update, I've played a few matches with texture quality reduced to medium and haven't had any further issues. To me this pretty much indicates that it is almost certainly the same (or a very similar) VRAM memory leak issue as being described in the post linked in my original post (as lowering the texture quality is suggested as a functional workaround for the issue there, too). It seems that this issue was somehow fixed on Linux, only for an eerily similar issue to show up in the Vulkan renderer on Windows. For now I'll play the game at medium texture quality until this issue might be resolved. I might test again on D3D11 to see if it also fixes the stuttering issue I was having there, to try and confirm if it might be the same issue manifesting in a different way on different renderers.