Odd performance behaviour around wheels

I noticed something, its always existed but the experimental build gave me access to the truck and its illustrated it in the most extreme way. I'm not sure what it is about wheels but they are REALLY resource intensive but it doesn't scale properly with adding more. Its like something is holding back the simulation to do with them. Its the t75 where it had the greatest impact. The wheels and tyres are costing too much in terms of resources and this really needs to be looked at. Even if as an FPS thing you look into an alternate mode of simulating them. Because this is silly. The t75 has a lot of physics lag associated with it, and it tanks my fps down to... odly its always 17-18fps regardless of the map. Its daft. My processor isn't the fastest, its an AMD phenom II x6 1090T. Its comparable to a low end AMD FX series 8 core insanitycessor. But otu of curiosity I wondered why two t75's had roughly the same FPS as one, I expected two to give me a slideshow rather than a game.

The t75, can you guess what my FPS are when it has no wheels? 90fps. I can have 5 no wheeled T75's and be happily simulating them at 40fps. The truck itself has a wonderful 90fps asociated with it. Gabester has actually done a great job of it, he could probably go mental if the wheels didn't cost so much in terms of resources. But its a very strange form of lag, one that suggests an actual bug or some horrendous inefficiency. One that might up the performance of the vehicles fantastically if solved.



But here's the weird bit of the behaviour. In the next screenshot I will have added one set of wheels, doesn't matter which be it front or back. See my FPS tank



From 90 to 22-25. Weird. and the physics became laggy. There are a lot of nodes, so maybe the lag is based on the amount of nodes. so I expect a slideshow upon putting them all back on.



And thats the problem. Its tanked all the way down to 17-18 fps. I lost 5 fps adding 6 more wheels, as opposed to losing 60 of them when I put on the first 4. The numbers are the same on any map, just more variable owing to what my gpu is rendering. But its the actual physics slowdown too, it renders a lot of things unenjoyable at times. What is it?

Bruckell Moonhawk with wheels - 35fps (some physics lag)
Bruckell moonhawk no wheels - 100fps

Grand marshal with wheels - 65fps
Grand marshal no wheels - 110fps

Cargo van with wheels - 60fps
Cargo van no wheels- 100fps

And the rest of the vehicles seem to perform well enough its no problem (although it certainly exists when I have fraps running and showing the FPS counter). This is a lineup of the worst performing vehicles on my system. Wheels steal about 40fps to 80 fps depending on the vehicle. Seems a bit much but I guess I can maybe forgive it. Just, each time the FPS drop is not even remotely linear. With the moonhawk for instance I'd expect to have about 60fps with just two wheels on. No, I lose all of that no matter how many wheels are on the car.

 

on smallgrid i also notice this, with all wheels and parts, 19-20fps with all parts but no wheels 50fps


EDIT:

i did some furthur testing, it seems to only affect the fps when you add either of the rear wheel sets (so either front rear or rear rear)


EDIT 2:

Took some screenshots (all with the front front wheels on)

Rear Rear Wheels on:



Front Rear Wheels on:



and no rear wheels at all:

 

Its a very odd bit of behaviour, and yes its a larger effect with the truck's rear wheels. In fact, front wheels knock me down the same old 40fps. Then I lose the other 20 when either back pair come on and another 3 when the next rear pair are attached. Wheels in general still eat too many frames in comparison to vehicles. If there's an argument for GPU physics simulation I think its for wheels specifically. But its odd Gabester doesn't have the problem at all.

I think wheels need some special paying attention to. Or even rethinking whether or not an alternative to node/beamning them is possible.

 

i only lose about 2-3 fps when i have the front wheels (50fps to 47fps)

 

Then half of this is probably my issue but I cannot think what on earth it could be. but its still a silly slowdown with the back wheels.

 

I'm getting this too. Goes from 80 FPS to 30-40 with rear wheels.

 

Something simmilar happens to me too (i5 haswell cpu). I've tested T-75 on grid map and I've got about 30 fps with all parts. But with removed one pair of rear wheels fps bumped to 120 fps. During removal I've noticed sth. weird, rear rear wheel was marked as rear front hub and front wheel tire. Maybe this is the cause.

 

Weird.. Turned V-Sync off, spawned T75; 116FPS, removed all wheels; 110FPS.

Tested with the other vehicles too, the difference isn't more than 10FPS, ever. I'll keep looking into it though, I have a laptop I can use for trying these things.

 

Should be noted that really the T series is a 10 wheel vehicle not 6. Each wheel in the duallies is individually represented which isnt going to be the most lag friendly setup.

 

I'll try this when I get home and post my results. It's weird how some people get it while others don't.

 

Logoster has an intel i5, says so in his signature and he to some extent experiences this. but this problem isn't global it seems. I get it for all vehicles, but the t75 is the worst for it. Its so strange.

 

that's another strange thing, while adding on one set of rear wheels causes a lot of fps loss, adding a second set only causes about 5 fps to be lost (at least for me)

there's some strange things going on with tires in the game it seems

 

Think about it though, the tyres are constantly in a collision state, the rest of the vehicle is not

 

It might be a problem with duallies. Both the H45 and T75 lag like crazy for me. I know they are the more complex vehicles, but considering the front wheels don't seem to do much to the fps, duallies could be the likely problem. I'll try to test this when I get home.

 

true, but it still shouldnt only cause huge fps drops with one set of rear wheels, if it's going to cause a huge fps drop, logically it should cause it per set of wheels, not with only one set of wheels

 

I guess the real way to test this would be to replace the duallies with just single wheels. See if that changes performance. Unfortunately I don't have the ability to test this myself.

 

i5 3570k 3.4GHz here. Here's some tests I did on puregrid at lowest settings to reduce my graphical bottleneck.
T75 no wheels: 200-270 fps
T75 front wheels: 210-260 fps
T75 front, front back wheels: 195-225 fps
T75 with all wheels and driving around: around 34fps (dips to around 23 fps on other maps and at highest settings)
T75 with all wheels while stationary: 80 fps (Not really relevant, but something interesting nonetheless. Has a form of node sleeping been implemented or is it just my CPU not doing as many operations relating to graphics?)

I get similar results with higher settings, but around 140 fps for the first three tests.

It's like the core that is simulating the vehicle reaches max capacity then craps itself and forgets how to breathe.

For reference:
Moonhawk all wheels: 200-250 fps
Moonhawk no wheels: 240-270 fps

Wheels are resource intensive because they use a damn lot of nodes and beams and the T75 has 10 of them with 16 rays each as opposed to the usual 12, meaning that each wheel has 66 nodes. Do the math. I can only imagine the T75 running perfectly well on systems with a very high tier CPU.

 

i5-2400 @ 3.6-3.7GHz, on small grid.

25 FPS - Full vehicle
92 FPS - No wheels
90 FPS - Full vehicle, no triangles
96 FPS - No wheels, no triangles

Hm. Wheel collisions? The 5th wheel collisions?