Optimal Settings for larger battles?

[PepperedSteakSlice29]

Any idea why my game is very laggy on predators in the mist?

I have a high end PC and I know the game is intensive for what it is, but it becomes exceptionally laggy on the pitched battles at the beginning of the DLC. I have even lowered the graphics settings to medium, but it still lags. Probably a CPU bottleneck, but if anyone has a solution please can you let me know as it is really frustrating. I would rather not have to turn down battle size if I can help it.

[Sunbather]

Reduce the fog (I don't know what the setting is called exactly), and also monitor closely what causes the low frames per second.
Is it only so and so many units? Is it a certain weather condition?

[case blau]

Aquire better CPU, then settings do not matter and you can play >2000 man large battles with no issues.

[Toni]

I somehow also got a lower framerate than usual when starting the first battle of predators in the mist (CPU 11900K). Do not know what caused it. I was thinking, it could be due to the same reasons that Sunbather mentioned. Allied AI units were participating in the battle, increasing the number of active units, and the weather was heavy rain.

But I did not do any tests, maybe it was just a bad constellation of several things hitting performance. A lot of smoke was present on the battlefield too = rendering smoke above foggy weather might lead to a lot of transparency calculations. So "reducing fog" like recommended may help - it generally is a helpful setting to increase performance in my experience.

[Zephyr]

For me all seems normal. My system does not seem to particularly like rocket artillery (like always) and certain artillery causes "hickups" sometimes with sudden frame drops. But that is all.

[Zyhgar The Phoenix]

This game doesn't really know how to handle modern systems very well or use their hardware at most I've seen this game use 40% of my gpu and draw 30fps which isn't that great.
Smoke and fog really bring down the frame rate a lot reason why rocket arty cooks a PC is the amount of smoke happening at once and probably it simulating fragments.
If you really want to squeeze some extra frames out of the game you could try using a custom DirectX API like DXVK I've tried it on my system gained 10 extra frames and it didn't show signs of glitching or bugging but I would say it is risky as DXVK is designed for linux and not windows so it may or may not work for you.

[Toni]

Originally posted by Zyhgar The Phoenix:

This game doesn't really know how to handle modern systems very well or use their hardware at most I've seen this game use 40% of my gpu and draw 30fps which isn't that great.

If the GPU only gets 30fps to render, then it will have a lot less to do than at 60fps or higher. So this is completely normal, most simulation games are CPU limited.

If your GPU would already be at, lets say, 90% when drawing 30fps then this would be bad, because you would have a GPU bottleneck then.

And even in a CPU bottleneck in a typical real-time simulation engine you will only have one core at 90+%. The overall use of the CPU will be much lower, due to many cores not being fully used.

So in your scenario with 40% GPU usage and 30fps: take a look at your CPU core usage. One should be above 85%...

But maybe we will see further parallelization in the future, like calculating terminal ballistics on separate cores or shifting more previously CPU determined processes to the GPU (by utilizing RTX features)? IIRC, then the GT engine already does "non-screen-image processing" (I mean the rendered image for the screen) related calculations on the GPU.

[Zyhgar The Phoenix]

I've only ever seen graviteam use 20% cpu load with multicore enabled and at best the cores it uses are 30-40% load while gpu is mostly 20-30% usage.

I'd do an performance experiment if i could find a performance tracking program that OBS can record as my current overlays don't show on OBS or steam.

And as I said I have used a custom API with about 10fps boost though this is with very little proper performance tracking and using API specific tracking which can be in accurate with frame timings.

Also "RTX" DXR is basically an very expensive lighting method RTX is just nvidias branding of it and turing cores specifically designed just for lighting calculations which will probably be replaced by Real time lighting which is cheaper. It has nothing to do with the gpu doing non image calculations as Async exists already in DX12 and VULKAN APIs and modern games.

https://steamcommunity.com/sharedfiles/filedetails/?id=2815777145 screenshot using API specific tracking using an DX9 VULKAN hybrid

[Zephyr]

Originally posted by Zyhgar The Phoenix:

I've only ever seen graviteam use 20% cpu load with multicore enabled and at best the cores it uses are 30-40% load while gpu is mostly 20-30% usage.

I'd do an performance experiment if i could find a performance tracking program that OBS can record as my current overlays don't show on OBS or steam.

And as I said I have used a custom API with about 10fps boost though this is with very little proper performance tracking and using API specific tracking which can be in accurate with frame timings.

Also "RTX" DXR is basically an very expensive lighting method RTX is just nvidias branding of it and turing cores specifically designed just for lighting calculations which will probably be replaced by Real time lighting which is cheaper. It has nothing to do with the gpu doing non image calculations as Async exists already in DX12 and VULKAN APIs and modern games.

New DLC with fighting :).
https://steamcommunity.com/sharedfiles/filedetails/?id=2824334424
https://steamcommunity.com/sharedfiles/filedetails/?id=2824334409

Small Battle in the snow: https://steamcommunity.com/sharedfiles/filedetails/?id=2824336517

It will give me anything between 40 and 70 FPs depending on the situation. Artillery (Rocket) may cause FPS drop still. With 6 BG or so I can also max out my RTX 3070 and will reach approx 80-90 FPS.

[Toni]

Originally posted by Zyhgar The Phoenix:

Also "RTX" DXR is basically an very expensive lighting method RTX is just nvidias branding of it and turing cores specifically designed just for lighting calculations which will probably be replaced by Real time lighting which is cheaper. It has nothing to do with the gpu doing non image calculations as Async exists already in DX12 and VULKAN APIs and modern games.

You can use it for much more than "lighting calculations". RTX multi-gpu workstations are heavily used for AI and deep learning (they use standard "game" RTX GPUs too). SB Pro (very similar engine) is thinking about maybe utilizing specific "RTX" features in the future too - they are not interested in lighting at all.

[Zyhgar The Phoenix]

This still does show that it doesn't properly know how to use hardware look at the CPU usage 100% a single core while the rest is on other random cores like that 40% on the 3rd core.
Modern games use even core multithreading ashes of singularity for example keep cores evenly loaded with 5~% error and properly use an GPU.
Also 7000mhz VRAM???????

[Zephyr]

Originally posted by Zyhgar The Phoenix:

This still does show that it doesn't properly know how to use hardware look at the CPU usage 100% a single core while the rest is on other random cores like that 40% on the 3rd core.
Modern games use even core multithreading ashes of singularity for example keep cores evenly loaded with 5~% error and properly use an GPU.
Also 7000mhz ram???? You running DDR5 ????

That is the GPU RAM (DDR6). And yes, Mius Front is a game that will run preferably on fewer cores with heavy loads, but it will also distribute among the others properly when that is not possible anymore. It will not give you the same performance while doing so. In Mius Front many things cannot be split between cores with equal effeciency it seems.

But again, it uses multiple cores and also rather well I think.

Just compare very heavy load in Korotich insane battle https://steamcommunity.com/sharedfiles/filedetails/?id=2824338085

with chilling in the snow
https://steamcommunity.com/sharedfiles/filedetails/?id=2824336517

[Zyhgar The Phoenix]

Originally posted by Toni:

Originally posted by Zyhgar The Phoenix:

Also "RTX" DXR is basically an very expensive lighting method RTX is just nvidias branding of it and turing cores specifically designed just for lighting calculations which will probably be replaced by Real time lighting which is cheaper. It has nothing to do with the gpu doing non image calculations as Async exists already in DX12 and VULKAN APIs and modern games.

You can use it for much more than "lighting calculations". RTX multi-gpu workstations are heavily used for AI and deep learning (they use standard "game" RTX GPUs too). SB Pro (very similar engine) is thinking about maybe utilizing specific "RTX" features in the future too - they are not interested in lighting at all.

DLSS is cool but AMDs FSR which is software level has managed to catch up and function on Non AMD gpus without needing fancy calculation dedicated cores or AI.

Originally posted by Zephyr:

Originally posted by Zyhgar The Phoenix:

This still does show that it doesn't properly know how to use hardware look at the CPU usage 100% a single core while the rest is on other random cores like that 40% on the 3rd core.
Modern games use even core multithreading ashes of singularity for example keep cores evenly loaded with 5~% error and properly use an GPU.
Also 7000mhz ram???? You running DDR5 ????

That is the GPU RAM. And yes, Mius Front is a game that will run preferably on fewer cores with heavy loads, but it will also distribute among the others properly when that is not possible anymore. It will not give you the same performance while doing so. In Mius Front many things cannot be split between cores with equal effeciency it seems.

But again, it uses multiple cores and also rather well I think.

It's clearly not using cores efficiently you have some cores doing nothing or actually probably doing background tasks for windows.
I feel they could get more performance out of it by tweaking the way the game talks to the hardware or even an API change though I've seen some impressive changes done to DX9 APIs to fix performance issues.
The fact the game feels smoother on an modified injected API designed for linux using VULKAN does clearly show there is room for improvement.

Also heres a slight difference between drivers
https://steamcommunity.com/sharedfiles/filedetails/?id=2247869222
https://steamcommunity.com/sharedfiles/filedetails/?id=2247869262
https://steamcommunity.com/sharedfiles/filedetails/?id=2247869200
As you can see driver over head is less eradic on more modern APIs

[Toni]

Originally posted by Zyhgar The Phoenix:

This still does show that it doesn't properly know how to use hardware look at the CPU usage 100% a single core while the rest is on other random cores like that 40% on the 3rd core.

Its the exact same in the only other similar milsim "SB Pro". Like said, they now try to get more stuff to other cores or the GPU (like terminal ballistics).
Graviteam has also constantly updated the engine for optimization in the background. It actually runs great given the insane detail of the simulation and the number of units in the tactical battles. Zephyrs screenshot shows it (I have to check usage on my PC too):
https://steamcommunity.com/sharedfiles/filedetails/?id=2824338085
And they do simulation related calculations on the GPU too, IIRC.

For the initial GPU complain:

Originally posted by Zyhgar The Phoenix:

This game doesn't really know how to handle modern systems very well or use their hardware at most I've seen this game use 40% of my gpu and draw 30fps which isn't that great.

GT will always fully utilize the GPU if needed. So in your "40% GPU usage, 30fps" scenario:

You can set an appropriate higher rendering resolution (super sampling with DSR or the AMD version) and get GT to use 90-100% of your GPU at 30fps while getting a much higher image quality at the same time. But it does not give more fps of course...

[Zyhgar The Phoenix]

I literally have a screen shot right there showing 40fps at 40% usage when i know the game can run at higher frame rates utilising my system to its full.

[Toni]

Originally posted by Zephyr:

Just compare very heavy load in Korotich insane battle https://steamcommunity.com/sharedfiles/filedetails/?id=2824338085

I have never had such a high usage % on the second most used core in GT. In your case 75% (first @ 92%). Too bad that you disabled total CPU usage. Would be interesting for comparison.

Can I ask: You turned off hyper-threading globally? Because it shows 8 cores. So turning hyper-threading off (I do not mean the setting in the game but turning it off in bios) is still a thing?