FSR Framegen?

[OnlyTaffingCowardsHide]

Any chance Amd's FSR Framegen and scaling could be implemented in TDM?
Seems it could solve a lot of problems with TDM being cpu limited on large draw distances, where if you can at least get 50-60 fps you can effectively get 90+ fps.
There was a lot of negative feedback surrounding framegen on release (and rightly so), but improvements have accumulated to the point where the benefits significantly outweigh any visible artifacting or input lag, especially on slower paced games like TDM.
I've tried Lossless Scalings framegen with initially good results, but I doubt it will ever be as good as FSR, and it currently struggles with dark scenes being a post processing effect. 
Maybe they could fix it if they offered post frame gen gamma correction in the LSFG tool, so you could set a higher brightness in game then lower it once the frame gen effect is applied.
Thanks to Optiscaler, once the work to add FSR has been done, there's no need to do any work to add DLSS or XESS as well, not that it would be possible to add those without paying for them.

Edited April 20 by OnlyTaffingCowardsHide

[Bikerdude]

Someone will correct me if wrong, but afaik, FSR/DLSS only works with DirectX and not OpenGL... 

Edited April 21 by Bikerdude

[lowenz]

You can port FSR 2, it's Open Source

[OnlyTaffingCowardsHide]

Might be something useful here: 
With Optiscaler the version doesn't really matter, we can use it to replace the FSR 2 with FSR 3/4, DLSS, and XESS. It should at least be viable on Linux by default, thanks to Mesa converting everything to Vulkan.

 

Edited April 21 by OnlyTaffingCowardsHide

[OnlyTaffingCowardsHide]

11 hours ago, lowenz said:

You can port FSR 2, it's Open Source

If I had the ability to do that, I wouldn't be trying to learn if it's even viable/begging others to do it...

[Bikerdude]

@OnlyTaffingCowardsHide

Whats are your full system specs, TDM will run on quite low end hardware.

[nbohr1more]

20 hours ago, OnlyTaffingCowardsHide said:

Might be something useful here: 
With Optiscaler the version doesn't really matter, we can use it to replace the FSR 2 with FSR 3/4, DLSS, and XESS. It should at least be viable on Linux by default, thanks to Mesa converting everything to Vulkan.

 

OpenGL 4.6 will mean that many players won't be able to use TDM. Our baseline is OpenGL 3.2.

Not that I'm against this but it would be similar to a lot of recent TDM changes where we tried to improve performance but the players who needed the most improvement had old hardware that couldn't use the new features.

Do you have a specific mission that is giving you performance problems?

Have you tried any of the "Performance Tweaks":

https://wiki.thedarkmod.com/index.php?title=Performance_Tweaks
 

?

 

[I.C.H.I.]

About performance. Make sure your GPU power management is set to maximum performance. A lot of fan missions suffer from the same issue when you traverse the level and driver is not quick enough to crank up the clock for heavier load, which creates very jarring framerate drop. The worst thing about it is that driver will downclock it back in the less performance heavy environment, and this creates pretty stupid ping-pong effect. The original FEAR also has this problem and even some modern poorly optimized games.

[lowenz]

FEAR has a performance problem due to.....input management: DirectInput FPS Fix - DLL files - PCGamingWiki PCGW Community

Unofficial Patch: Wemino/EchoPatch: Modernizes F.E.A.R. and its expansions with HUD scaling, high-framerate optimizations, controller support, and other quality-of-life enhancements

Edited April 23 by lowenz

[I.C.H.I.]

4 hours ago, lowenz said:

FEAR has a performance problem due to.....input management

No, this happens even in the fully patched game with EchoPatch. This is not an exclusive issue for FEAR, the driver is to blame. Dolphin developers always recommend to enable maximum performance to have smooth framerate, and PCSX2 also needs it.

[chakkman]

TDM performs pretty good IMO. Sure, the engine has its limitations (as far as I know, especially outdoor areas have to be arranged carefully to not become too demanding), but, all in all, especially with the updates which improved performance, and the multicore optimizations, it's running pretty well now. No complaints here, and, I don't have a high end rig (RTX 3060 Ti and a i5-11600K).

In my experience, especially antialiasing set to 16x was a massive hog. If you use 8x, which is more than enough, TDM runs well.

[I.C.H.I.]

1 hour ago, chakkman said:

antialiasing set to 16x

IMO 16x needs to be hidden by default and available only through the console or TDM should warn people about huge performance impact.

[Arcturus]

18 hours ago, nbohr1more said:

Our baseline is OpenGL 3.2.

Just checked and OpenGL 3.3 came out 16 years ago.

[OnlyTaffingCowardsHide]

19 hours ago, I.C.H.I. said:

About performance. Make sure your GPU power management is set to maximum performance. A lot of fan missions suffer from the same issue when you traverse the level and driver is not quick enough to crank up the clock for heavier load, which creates very jarring framerate drop. The worst thing about it is that driver will downclock it back in the less performance heavy environment, and this creates pretty stupid ping-pong effect. The original FEAR also has this problem and even some modern poorly optimized games.

Like laptop power settings? Or disabling low powerstates in overclocking software like Afterburner?
The game seems to be running alright gpu side (rx6600).
Shadow maps can be intensive, and even stencil shadows can run poorly for me if cranked up (ryzen 4500 cpu). 
It is a little strange that we can't get soft edged shadows without sacrificing too much performance,
"fast approximate soft shadows" would be a neat feature for low end gpus, that just make the edges soft for minimal performance hit...
One problem for me was getting the game to sync properly on a 60hz display (plasma).
Logic would tell you that enabling vsync would enable perfect 60fps frame delivery, but I actually have to uncap the fps, then set it to 60. I don't disable vsync, but something is wonky with the default framerate cap that makes it impossible to maintain a smooth 60 without choppiness. 
Beyond that I think either my cpu is just too weak to handle long draws, or the engine is just never going to be cpu optimized. I wouldn't expect it to be, given that it's probably only using 1 core, and likely doesn't care about modern instruction sets or cache, but it has become more of a problem as mappers push the engine. I've never been able to set lod's above "normal" for that reason. If anyone is able to use the higher settings without stuttering, I'd love to hear about it (what cpu are you using?). That's where I think framegen would be most useful, not just for weak gpus but for engine cpu limitations in general.

Edited April 23 by OnlyTaffingCowardsHide

[OnlyTaffingCowardsHide]

5 hours ago, chakkman said:

TDM performs pretty good IMO. Sure, the engine has its limitations (as far as I know, especially outdoor areas have to be arranged carefully to not become too demanding), but, all in all, especially with the updates which improved performance, and the multicore optimizations, it's running pretty well now. No complaints here, and, I don't have a high end rig (RTX 3060 Ti and a i5-11600K).

In my experience, especially antialiasing set to 16x was a massive hog. If you use 8x, which is more than enough, TDM runs well.

I'm not sure it does though. In my experience performance is inconsistent. 
It can perform well if I set everything to low and disable soft shadows. I've had it running on a phone from 2021 that way and was quite impressed that it even ran let alone at 60+ fps.
When you start turning things up though, fps drops dramatically. MSAA isn't the end of the world for me, I'm working on a reshade preset that allows SMAA and a form of TAA that should be superior to Unreal's pitiful implementation... so I don't even use msaa above 2x or 4x... 
It's the lod setting that gets me every time I try to set it above normal.
The only other thing that kills performance for me is shadow maps, and that's likely due to AMD's crappy windows drivers. I had better performance on a 1660 Super.
If LOD is placing a heavy load on the cpu I'm not sure there is any way around it, if the game can't utilize more than 1 core, we're kinda stuck for the foreseeable future. I'm sure an extra 800mhz core clock would make a difference for me, maybe l3 cache is relevant in this game as well?
I can live with the popin from LOD, but it's not ideal compared to modern games.
Implementing SMAA engine side probably wouldn't be a bad idea though, it'll likely run even better than opengl reshade variants. Marty's Immerse SMAA is somewhere between MSAA 4x and 8x, and runs similarly to MSAA 2x, so there's no reason not to use SMAA these days.

Edited April 23 by OnlyTaffingCowardsHide

[Bikerdude]

On 4/22/2026 at 11:44 AM, Bikerdude said:

@OnlyTaffingCowardsHide

Whats are your full system specs, TDM will run on quite low end hardware.

Bump.

[I.C.H.I.]

1 hour ago, OnlyTaffingCowardsHide said:

Like laptop power settings?

I don't know how it works for Radeon but for NVidia there is a setting in the control panel called Power management mode, it's is also possible to change it for any game individually. Also Windows has its own power management settings. Here is the thread on Dolphin forum about it(but it's pretty old and some things most probably changed): https://forums.dolphin-emu.org/Thread-unofficial-laptop-performance-guide

[OnlyTaffingCowardsHide]

2 hours ago, Bikerdude said:

Bump.

Sorry buried my specs in my long drawn out posts...

Ryzen 4500 6 Core (main weakness is 8mb l3 cache)
16gb DDR4 3200
AMD RX 6600 (Upgraded from 1660 Super, and Rx 590 before that, noticed no benefit to lod setting nor shadow maps).
NVME SSD 

Like I said in my previous posts. I would think if it was gpu related I'd notice some benefit on the rx6600 vs the rx 590, but the only thing I've noticed is that the 1660 super did well with shadow maps, and reshade compared to the rx590, whereas the rx6600 is roughly on par with the 1660 super in opengl games, despite destroying it in DX/Vulkan.
So I'm guessing it's either the 4.2ghz clock, or the 8mb l3 cache holding things back for the lod setting. I can't imagine an extra ghz would make a world of difference here, but maybe TDM benefits from l3 cache.
I'd be happy to hear that others have had better luck than me setting lod higher than "normal", maybe the engine is just not capable of utilizing hardware to push that setting.

Edited April 23 by OnlyTaffingCowardsHide

[OnlyTaffingCowardsHide]

2 hours ago, I.C.H.I. said:

I don't know how it works for Radeon but for NVidia there is a setting in the control panel called Power management mode, it's is also possible to change it for any game individually. Also Windows has its own power management settings. Here is the thread on Dolphin forum about it(but it's pretty old and some things most probably changed): https://forums.dolphin-emu.org/Thread-unofficial-laptop-performance-guide

Alright, windows is on default full power, and I set a custom power usage profile in the AMD control panel to use up to 120% power... no difference. 

[OnlyTaffingCowardsHide]

Turns out it is gpu related... the game just doesn't use more than 70% of my gpu without stuttering... neat. Maybe it's just a windows/AMD/Opengl problem. Really thought it was a cpu related draw distance problem, as that would normally cause low gpu usage, but disabling all the effects except for lod on max completely solved the problem. Guess you learn something new every day.
So I guess I need a better gpu, or a linux install, or a lighter reshade setup.

After extensive testing I can confirm that cpu usage is actually quite good. The game can use around %50 of my cpu at a locked 60. When I try to go above 70% gpu usage the framerate drops and cpu usage drops, which is expected gpu limited behavior... 
 

EDIT:

I'm a m0r0n, had msaa enabled while testing maps vs stencil shadows, can confirm stencil shadows are much more efficient. 
Turned on SMAA instead at a cost of only 4% vs MSAA 4x gobbling up 14% gpu usage.
Now I can have soft stencil shadows, volumetrics, SMAA, and MXAO running all at once thanks to reshade. Which would have been impossible without it thanks to the gpu usage cap. 
So I guess framegen would be useful for many people running AMD cards afterall, not just low end/older rigs.

Edited April 23 by OnlyTaffingCowardsHide

[jaxa]

23 hours ago, I.C.H.I. said:

About performance. Make sure your GPU power management is set to maximum performance. A lot of fan missions suffer from the same issue when you traverse the level and driver is not quick enough to crank up the clock for heavier load, which creates very jarring framerate drop. The worst thing about it is that driver will downclock it back in the less performance heavy environment, and this creates pretty stupid ping-pong effect. The original FEAR also has this problem and even some modern poorly optimized games.

I wonder if I'm running into that with an RX 560. Obviously not very fast, but I ended up capping the frames at 30 for some missions like A Bridge Too Far and The Wizard's Treasure. At 720p.

I need to enable an overlay so I can check the GPU utilization.

[OnlyTaffingCowardsHide]

So to sum up, msaa sucks, shadows maps are gpu heavy, this game can't use more than 70% of AMD gpus running Windows, it had nothing to do with power settings or cpu usage, and I can happily play this game now without stuttering.
Most people will probably just want to use Linux for AMD cards, but I like using reshade so I'll live with it.
Framegen would be more useful than I thought, even for midrange AMD cards.

[OnlyTaffingCowardsHide]

37 minutes ago, jaxa said:

I wonder if I'm running into that with an RX 560. Obviously not very fast, but I ended up capping the frames at 30 for some missions like A Bridge Too Far and The Wizard's Treasure. At 720p.

I need to enable an overlay so I can check the GPU utilization.

Probably not, the RX 560 is a truly pitiful card... User bench has it 160% behind the RX 590.
Linux might help, but learning to swap power supplies is a very useful skill. There just aren't any decent affordable gpus that can run on motherboard power alone.
Some missions are just really poorly optimized. I haven't been able to play A bridge too far on any of my gpus. 

Edited April 23 by OnlyTaffingCowardsHide

[Bikerdude]

On 4/23/2026 at 8:17 PM, OnlyTaffingCowardsHide said:

• Ryzen 4500 6 Core
• 16gb DDR4 3200
• AMD RX 6600
• NVME SSD 

This is more than fast enough to play TDM @1440p in medium to high settings.

[stgatilov]

On 4/23/2026 at 8:17 PM, OnlyTaffingCowardsHide said:

Ryzen 4500 6 Core (main weakness is 8mb l3 cache)
AMD RX 6600 (Upgraded from 1660 Super, and Rx 590 before that, noticed no benefit to lod setting nor shadow maps).

I used to have Ryzen 1600 before the upgrade and always thought it was above what a typical TDM player has.
I have switched from GeForce 1050 ti (which was perfectly OK for TDM) to RX 6600 and I don't see any performance issues. I just set LOD to maximum because I hate popping and LODs in TDM are usually light-weight enough to just tank through them on my machine.

Either you use some very heavy settings (insane antialiasing, maybe 4K resolution, etc.) or you expect 300 FPS, or there is something wrong with your machine.

On 4/23/2026 at 4:39 PM, OnlyTaffingCowardsHide said:

It is a little strange that we can't get soft edged shadows without sacrificing too much performance,
"fast approximate soft shadows" would be a neat feature for low end gpus, that just make the edges soft for minimal performance hit...

Everything related to shadow maps is quite bad in the game, unfortunately.
I have been trying to rework it completely lately...

Quote

Logic would tell you that enabling vsync would enable perfect 60fps frame delivery, but I actually have to uncap the fps, then set it to 60. I don't disable vsync, but something is wonky with the default framerate cap that makes it impossible to maintain a smooth 60 without choppiness. 

The game modelling from the original Doom 3 works with fixed 16 ms timesteps, which cannot give you smooth framerate even with VSync. The right way to go is "Uncapped FPS" which means using dynamic timestep in game modelling (I believe the name of this setting is just bad).

In fact, we have made the "Uncapped FPS" default a few releases ago, so that people don't suffer constant stutters with default settings.

 

If you want, you can record some hundreds frames of gameplay with Tracy (instructions here). Then you'll at least see what limits your case. You can send the trace to me as well.
And please post your darkmod.cfg.