Odd performance behaviour around wheels

hmmm, just noticed that the only 2 people who don't have this problem are both developers(at least that's posted there results so far) (gabester and drowsysam)

maybe there's something in the developers build fixing the fps problem that's not in the experimental build? i have no idea

 

Or maybe they have nice intel CPUs.

I tried to replicate this problem. Grid empty map, everything maxed out 1080.

FPS with all wheels 72-73

FPS with no wheels 72-74

Pretty much the same.

 

FPS with wheels 15
FPS without wheels ~35-45

 

Same here. I7 2600, GTX570

All wheels 23fps, without rear rear wheels 70+ fps

 

@Hati

When you have the t75 with all wheels, can you try disabling the UI (Shift + U) and see if there is any difference in FPS.

Edit: Because the UI eats 14-15 fps in my case when I spawn two t75, so that means the UI likes to use quite a lot of CPU resources.

With UI: 61 fps


Without UI: 75 FPS


Without the UI I have the same FPS with 1 or 2 t75.

 

Stock T75 and I get 20-25FPS with my FX-8320 at 4.8GHz. If I remove either set of rear tires I get about 115FPS.

 

It made zero difference. What I have along side my problem it is actual physics lag too. Its all running at around 75% realtime until the wheels go away. But someone here brought up wheel collisions. There's one way to find out I guess. I have to wonder at something though, its all of us who don't have top of the line CPU's who get it to some degree. However I have enough cores to handle those wheels multiple times over. Do you think we're running into a wall set by clock speed? because from what I see on my performance monitor the game kinda forces the simulation of the truck onto one thread. All six of my threads get a minor rise out of the truck but its all handled in one place basically. They don't share the load (which in of itself may help with simulation speeds). So a higher clock speed would reduce the problem drastically (because it is peaking). but ti just means the current code doesn't make full use of the resources my computer has to give.

However, someone might have hit on something with self collision of wheels. I disabled it by editing the files and while I still have the abysmal FPS I have no more physics lag. Still only 20 frames per second but I have no more physics slowdown. I can play in realtime

 

I don't really get that much of a performance increase without the wheels. It's in the realm of 8-10 FPS difference, as illustrated below. AMD FX-6100, currently stock (3.3 GHz) after an overclocking fiasco reset my bios, so the clock in my sig is no longer accurate. I guess it's a simple matter of a CPU upgrade for me?

Full Wheels


No front rears


No rears


No wheels


See? Not really all that much of an increase, but still noticeable.

 

I made a comparison of CPU usage with the d15 vs t75 vs 2xt75

 

Same deal for me, just tried it. I run Vsync with a 60fps frame limiter, the framerate gets cut almost in half with the T75 but goes back to 60 when I remove both sets of rear wheels. Removing only the Rear Rear set seems to help more than removing only the Front Rear set. It seems to be purely a physics issue, if I play in slow motion mode with alt+down arrow it smooths out to 60 again. Closing the UI does nothing.

I'm on an i7 4700MQ that usually has no trouble with physics calculations. Interesting how this destroys some peoples framerates but has no effect at all for others.

 

That's because vsync limits your framerate to even divisions of your monitor's refresh rate. Can't make 60? Only 58? Run at 30. Can't make 30? Only 28? Run At 15. It should always be turned off when testing for FPS.

 

This is a really odd problem.

Max settings, 1080p and UI off

With the T75 with wheels I get around 20 fps
Without the wheels that is around 75-80 fps. Exactly the same as what I get when I turn off my physics

This is really bad towards any hope of fixing the current wheel model too if the performance is so low with only these basic wheels...

 

No, that's not the issue. Triple buffering is specifically designed to stop that from happening, I have that forced on in the Nvidia control panel.
It's definitely a physics thing, normal slowdowns on complex maps like Dry Rock will drop the game down to about 45-50, but the T75 wrecks the framerate no matter what map unless I remove the wheels.

 

Just out of curiosity, do you get a physics slowdown? the difference between dropping the physics to 2x slower than realtime and setting it to realtime for me is hardly there. If you get the physics lag then disable the self collision (as per this file). For me it eliminates the lag but my FPS is still in the sub 20's.

 

All wheels: 24fps
Front rear wheels missing: 70fps
Front wheels only 94fps
All back wheels: 26fps

So it seems that the problem is also with me :/

i5-2500k processor at stock clocks.

 

I'll test later on my brother's i3-2120

 

BeamNG simulates each vehicle in its own thread. 1 thread per active vehicle. 1 Thread can only ever run on 1 core at a time so it doesnt matter how many cores you have, the T75 will be crammed onto one. If the other cores get a minor usage bump it means the T75 has completely taken over 1 core and all the other tasks also using that core have had to be offloaded elsewhere. So you have hit the single core performance ceiling for your CPU which is why in all the Intel vs AMD debates on this forum you'll find a ton of recommendations for Intel as their single core performance at the same clock speed as AMD is higher.

- - - Updated - - -

With my settings on grid pure
1 D15, all wheels: 24 fps
1 D15, front wheels removed: 40 fps although it is erratic and sometimes drops to 24 again (I also had it reach 50 once)
1 D15, no wheels: sits nicely at 60 fps


1 T75, all wheels: 7 fps and noticeable physics slowdown
Remove the rear rear wheels: 14 fps
Rear rear + front rear removed: 17 fps
All wheels removed: 18 fps
If I then remove the exhaust stacks, fuel tanks and mudflaps I hit 40 fps so there is clearly some slowdown associated with those too.
Re-add the wheels with those extra parts removed: 14 fps

 

Core i7-3770K also does this. 25fps with all wheels and 79fps with 1 rear axle removed .... but i also had the effect that the game partially froze .... no physics anymore but i could cliick around in the UI.

 

Thats why I removed the wheels themselves instead of the entire axle.

 

Ok now this is interesting, removing wheels made me lose FPS

All wheels:


Front removed:


Fron and rear removed:


No wheels: