Turning off vsync helped getting rid of the massive input lag and "pumping"(switched from fluid to laggy in seconds) when moving.
Due to how vsync works in can half your framerate(more detailed explanation here:
http://www.tweakguides.com/Graphics_9.html ).
Turning on vsync with triplebuffering and prerendered frames (via the nvidia control panel / catalyst control center) i now get a stable framerate - it is not higher - just stable and i do not suffer from that horrible input lag i had in the beginning.
If your computer struggles to get a high framerate in the first place the above may not solve this.
I tried to play Log2 on my old AMD 64 3200+ with my old ATI/AMD Radeon 3850HD AGP card and it was almost unplayable :/ (even more then MMX legacy) If this would have been a turn based game i wouldn't mind it but since it is realtime your framerate matters. There was not much going on graphics wise to explain why the framerate was as low as 15-20. Yes this system is old but i can see in most games why the framerate is bad, but i cannot see why Grimrock performs so bad - even if it looks nice.
Maybe some things can be compiled / hardcoded in a map file so that the cpu/gpu does not have to do everything during runtime, but then the editor wouldn't be so powerful and flexible as it is now(you can change everything on th fly and try it out instantly) without having to compile everything and get a coffee or 2 while waiting for that to be finished - only to see that you forgot something or something is broken... but at least that was a fast method.
Or are the scripts the issue here? Seeing this post:
http://www.grimrock.net/2014/03/07/perf ... mizations/
There is an "unclassified" process which used 52% of the ressources. Whatever it is or what it does it seems pretty ressource heavy compared to the other processes.
I still go with the idea of a huge static mesh which gets blended in/out by the lod system, or more 2d objects which gets lodded to their 3d counterparts - but this looks pretty ugly and requires a non trivial amount of extra work just to make slower systems run the game.
Can more things be parallelised so that multi cpu systems are able to profit from their extra computing power? It doesn't seem so right now.
PS.: What is the console command to show the triangles? I guess r_showtris 1 only works for quake ;b