This is a great game but certainly need serious improvement ,mainly for people who have powerful GPUs but lesser CPUs and if I'm not wrong that could be done using GPU acceleration to improve linear algebra operations https://developer.nvidia.com/gpu-accelerated-libraries
Wouldn't then the performance be dependant on what GPU brand you have? the game is less CPU heavy than i thought. Maps are entirely GPU based, i think shadows and lightings are CPU, and multiple cars and car deformation is CPU based.
From my experience in this game when you have a r9 nano and a FX 6300 it doesn't make much difference between the best and worst graphics and from my knowledge of hardware the shadows and lighting are processed by the GPU
The game is very well balanced, it is your system that is not balanced and the reason for the performance issues. You have an extremely good GPU and a poor CPU.
Shadows are most definitely using CPU power and a lots of it, albeit only on single thread, 10-15% on my CPU, of course depends a lot from number of objects and amount of lights. Next update will have optimizations, what optimizations is not exactly known yet, but something. With weak CPU your only hope is to keep lights off and shadows off, also there are other graphics settings that are helpful to reduce with low CPU power, mesh detail from high to normal helps some for example, keep dynamic shadows off too. Depending what you do in BeamNG, you might need more single core performance in current version than multi core performance to keep that GPU happy. Overclocked newest model I5 might be faster than CPU with more cores, it depends what you do in BeamNG and nobody but devs knows how that will change in 0.9. Problem with FX CPU is that it does not like BeamNG, but I would guess without shadows and dynamic reflections you won't have performance issues, but of course without shadows many will not like to play it.
But this way you could reduce the demand on the CPU if you have a powerful GPU --- Post updated --- Nope already tried that its a 5 fps difference and the FX is not good at floating point calculations that's what make the game almost unplayable with it mostly because its CU is more used in textures and mesh than the ALU that is what make beamng lags on FX --- Post updated --- The post is about using GPU acceleration to make systems like mine run the game better with NO HARDWARE UPGRADE not about if my system can or not run the game that i already know
I have impression that not many really know how to actually test these things, do exactly same settings as in video, same way disable shadows, with your more powerful GPU you should see lot more fps boost. Most definitely shadows are done by CPU, you can see in video how I'm CPU limited with shadows and GPU limited without shadows. Using that many lights in car of course exaggerates the effect, but that is kinda the point. Shadows can be quite easy on CPU on some places, especially if not using lights, each light multiplies number of shadows, each object creates own shadows, hence big impact on video. Now if physics are also maxing out your CPU, then when you have shadows off, press J and you should see how much shadows really affect, because when you hit one limit, any limit on CPU, there will be waiting on GPU, monitoring GPU load and having something to monitor CPU peak load as those many needles on top of the video will reveal what is the truth. That East Coast town with that car config is probably heaviest in terms of shadows that there is in vanilla content. With monster GPU I would be easily led to believe that any CPU available will peak at 100% with single car in this test. Faster GPU needs faster CPU to feed it if goal is maximize GPU usage and that goal would be good with monster GPU. What I have found out is that for some reason, task manager can't keep up with fast peaking CPU load of BeamNG graphics thread, so I have to make my own CPU monitor to see what really is happening. Those needles show peak during 250ms, 50ms interval of reading CPU load and highest from last 250ms is shown by needle, so I know when my CPU is maxing out much better than any other CPU load monitor. What happens when core running graphics thread maxes out is that GPU has to wait, so GPU load drops down, so any time GPU load is going to be less than 99-100% CPU has peaked out, I believe if core running physics thread maxes out, then again graphics have to wait. With FX, solely based on what I have read from these boards, it appears to me that it is physics that limits fps. You can tell which core is running physics by pressing J-key, which core drops a lot in load is one running physics. If that maxes out, there is really nothing that would help you except changing CPU. I doubt that there is much interest to code workarounds for FX cpu as it is mostly the problem here, at least that is my guess.
