Misery 2.1.1 DX10 & 11 Memory Leak Thread - MISERY mod for S.T.A.L.K.E.R.: Call of Pripyat

Hi all,

Misery 2.1.1 definitely has a DX10 & 11 memory leak. When first loading the game on DX10 or 11 I have ~120 fps with maxed out settings, however, after ~10 minutes my fps is ~60 and after 30 minutes my fps crawls to ~30 fps.

I have a brand new ASUS ROG STRIX GTX 1060, i5-4690, 8GB RAM and the game/mod is installed on my SSD. All my drivers are up to date and I max out modern AAA titles at 1080p with ~100 fps no problem. Why is this happening?

I have read a few threads discussing this already, here on ModDB and elsewhere, but no one seems to have found a solution.

I have tried core affinity tweaking, disabling AA, Tessellation, Volumetric Smoke, Sunrays, different resolutions, you name it, but nothing works.

Fortunately, I can play the game just fine maxed out on DX9 Full Enhanced Dynamic Lighting at ~150 fps without the memory leak but I miss out on the pretty water/wet effects, volumetric smoke and tessellation

Does anyone have any further ideas I could try?

Thanks for reading and allowing me to vent a little

LeNimble

I have played 5679 hours on dozens of system builds with Nvidia and AMD cards on every renderer. I primarily play on DX11 except when testing I use static for quicker loading to try things out. I have never experienced decreasing FPS. There is a bug (fixed for 2.2) when you accept but have not yet completed the hungry merc mission. The system will constantly check to see if you have enough food and bogs its self down. I always just kill the arrogant bastards, but we fixed it for 2.2. There is a problem with certain tree models causing performance hits which is fixed in AN4.

Other than that, you can try both the latest and older versions of your drivers. The problem will lie with the GPU drivers generally. COP is not a modern title, expecting the same performance on an eight year old engine is unreasonable TBH. The video card manufacturers don't concern themselves with ancient games when they build their drivers. I easily pull mid 100's FPS, but limit at 60 because more is simply a waste of energy and more heat. At 60 FPS my GPU fans never even need to run. Limiting also greatly reduces stutter and lag, even on an SSD.

Hmmm, I will try vsync but I'm sure the slow down is CPU/RAM bottlenecks. I have monitored my GPU usage and the 1060 is overkill for CoP/Misery, as you said, the fans don't even turn on and it's not stressed at all. It's not the hungry merc mission, I never even spoke to them and gunned them down many hours ago in this play through. I'm currently end game in Pripyat, about to do Lab X8. I originally installed AN 3 Lite over Misery 2.1.1 for this play through to have a greener zone but I don't think it's the trees causing this as they would be no different in DX9? I'm sure it's something to do with the effects enabled via DX10 & 11 which leak out

If it is a driver issue that sucks as it means i'll never be able to play Misery with DX10/11

I will do some more tests tonight after work. I'll try DX11, vsync (well, adaptive sync), and turn off all the effects DX10&11 actually introduces, so water effects, volumetric smoke and tessellation and see what happens. I'll also try with/without sun rays to see if that's causing it but it doesn't in DX9. If it is the sunrays i'll go with DX9 as I need those more than water effects

Any other ideas?

My 1080 probably uses the exact same drivers and I have no issues. I would try it without AN3 lite, just vanilla game plus 2.1.1.

Besides mutant AI, the following are the worst performance killers in order of priority: AA, Shadows, Sun Rays, and Dynamic Lighting.

I never use Sun Rays personally. I also never use vsync. I use dxtory for frame limiting.

You need to use a particular method when changing video options to ensure they are really applied. I have thorough instructions in an article in my comments section.

Once you know the procedure, it takes less than 30 seconds each new combination you try. It saves a ton of confusion.

Please report your findings here.

p.s. I also verified through exhaustive testing that HDAO (I use the low level) works just as good with Nvidia as AMD cards (I have had dozens) even though it was developed for AMD brand. The graphics are better on HDAO than HBAO too. It seems to use your GPU to better effect and unloads the CPU some.

Okay thanks, yea I have read a lot of your posts and have already tried new user.ltx files etc. I use HBAO generally but will try HDAO tonight.

I'm not sure I can uninstall AN3 Lite and don't want to risk corrupting the mod/saves at this point. I'm only a few hours from completing Misery 2.1.1 (FINALLY! ). So once it's over I will reinstall Misery completely and try AN 4.

Will report back here later tonight. Thanks again

AN4 is a work in progress. We are not sure we have worked all the bugs out yet. All you would need to do is backup your saves. Reinstalling if all you change is AN3 will not effect them. To reinstall, you need to uninstall Misery, uninstall the game, delete all folders (backup your saves first), reinstall COP, start and save once, install 2.1.1 FULL version as admin and then play with your video settings.

You can, of course, finish the game first. I would backup a save late in the game where you are experiencing problems to use for testing.

Right, early testing and report here but I may have solved it or at least improved it somewhat but I would need to play for longer to confirm (damn life responsibilities getting in the way ).

Anway, I reset the graphics options to DX11 maxed out, minus AA and Tessellation, with adaptive sync forced via the nVidia CP. For the hour I played my fps remained at 60 except for the stutters and a brief drop into the 50's when it was stormy, raining and there was lots of grass and trees in my FOV (acceptable). I never use the CoP AA as I run the game in 1440p DSR on my 1080p monitor, so it can't be the AA that altered the performance. It's either the adaptive sync, no tessellation or both. Seeing as DX10 doesn't have tessellation but also sees performance drops, my testing suggests it may be the adaptive sync. I have no idea why this would be the case and if anyone could explain this to me I would be very appreciative!

Will do further tests over the weekend to confirm

What is adaptive sync? Why 1440p?

Adaptive sync is just a better form of vysnc by nVdida, basically if goes below 60, instead of cutting straight to 30 (causing jerkiness) it allows normal frame rate anywhere between 30 and 60, basically vysnc cap at 60fps. 1440p DSR or Dynamic Super Resolution is another great nVidia feature whereby it processes the game in a higher resolution than ones monitor (1440p in my case) then downsamples it (with a smoothness filter) to my navtive 1080p resolution. Effectively super sampling. It's an alternative to AA but taxing. I don't use it often but AA is CoP is terrible and it's and old game so I can run DSR no problem.

What is your monitors maximum refresh rate? Frame rates above that will cause it harm and degrade the picture. Just out of curiosity, can you show me side by side screen shots of the same scene at 1080P without adaptive sync in use and with it? I can use any frame rate I choose from 30 to 150 with dxtory without using any additional processor or GPU power.

My monitor is 60hz. I don't usually use any form of vsync as I hate input lag but thought I would test it in this instance anyway and it seems it may pay off. I use the nVidia Shadowplay fps counter to mointor my fps in all games to ensure the sweet 60+ / ideally100 fps spot, if not more, so that I never dip below 60. With adaptive vsync it actually plays the exact fps below 60, so 51 fps will play as 51. If you use normal vysnc, you may see 51fps with a certain fps monitors but you are actually seeing a 30fps lock to avoid any screen tear. Will put up pics over the weekend if you want

I never use vsync at all. I limit at 30 or 60 FPS. It stays within a frame or two of that except for the odd dip as new alife comes online, usually very short. I always like to compare side by side pictures. I had to do it to notice the differences between HBAO and HDAO.

You won't be able to see anything from screenshots with any sort of sync options. Tearing happens on the display and is not capturable by a standard screenshot.

I use Nvidia's driver a-sync as well. If you're someone who keeps settings low enough(or has a beefy enough PC) to keep frame latency low it's pretty effective at removing tearing but it also doesn't kill your performance when you drop below your refresh rate just a little. Of course if you have a 120hz display it's probably not as great.

Now if Nvidia supported the vesa a-sync standard that would be great, but they're a bunch of fucks so it won't happen.

I finished the game/mod over the last weekend (YAY!) and noticed the fps drop again, but after an hour or two. So a-sync and no tessellation seem to help a bit and keep the frames up for longer but it eventually crawled to ~30 fps :/

Anyway, I'm moving onto the new Deus Ex and other games until Misery 2.2 so will wait for that and see how it goes, but thanks for all the help/replies. Good hunting, Stalkers.