MX Simulator

Loaded up 100 of them in a blank track and it was totally smooth. The 2048x1024 completely black spec map should be scaled down to 16x16 but aside from that looks good to me.

Maybe for Washougal there was a bad collision file that wasn't in the tree pack. Getting inside the bounding sphere might have coincided with the LOD swap leading you to think it was the LOD.

jlv wrote:The highest level of detail doesn't really matter as far as frames/sec goes. 10,000 vs 11,000 doesn't matter much. The problem is the lowest level of detail. There it's 10,000 vs 500. That's a killer. It leaves the game engine with no quick way to draw the model.

This is good to know and it could be the reason some of the lods I've used caused issues with skipping as they come into view. What do you think is a good step per lod? 10%, 20%, etc.

aeffertz wrote:

jlv wrote:The highest level of detail doesn't really matter as far as frames/sec goes. 10,000 vs 11,000 doesn't matter much. The problem is the lowest level of detail. There it's 10,000 vs 500. That's a killer. It leaves the game engine with no quick way to draw the model.

This is good to know and it could be the reason some of the lods I've used caused issues with skipping as they come into view. What do you think is a good step per lod? 10%, 20%, etc.

Each LOD is double the distance of the previous LOD (assuming "model geometric detail" is at 50). Since the distance is double, the area on screen is 1/4, so each step should ideally have 1/4 the geometry. That's in theory. In practice that will be almost impossible to achieve.

That tutorial I linked made a good point about doing test renders at the approximate resolution your model will appear at in-game. (Their LOD numbers are different though. Their LOD 2 would be more like LOD 3 or 4 in MXS which is generally where I'd start.) So for LOD 4, set up your camera's FOV to match the game, move the camera 16 feet away and do a 1080p render and see what you can reduce. For LOD 5 move it 32 feet away. Your model will probably be ~100 pixels high at this point so you'll be able to remove a lot. For LOD 6 it'll be 64 feet away and only ~50 pixels so you'll probably be able to really butcher it at that point without being noticeable.

***Bit of a mess this post***

Just a bit of a test with my shitty laptop. Not much difference with framerate until I bring the model geo detail down to ~5. Not sure if the frame rate step up just comes from bringing the clipping plane forward, but the model polygon count didn't really change significantly. The reason I took away alpha channels-the original angle I was going with this post was that perhaps the overlapping alpha channels of all the trees being rendered over one another may have been the culprit. But that doesn't appear to be the case.

Model Geo Detail 50%


Model Geo Detail 50%, but with no alpha, no billboards, no collisions


Model Geo Detail 15%, but with no alpha, no billboards, no collisions


Model Geo Detail 5%, but with no alpha, no billboards, no collisions



____

Further too above:



Model Geo Detail 100%


Model Geo Detail 100%, but slider reduced to bring forward model clipping plane. This one is interesting, as it is rendering more polies than most of my other screenshots, but frame rate is better

I think you might be fill rate limited. Does it speed up when you reduce the size of the window?

I will definitely look into making lods for all my models. I think for me I just needed a bit of clarity of how they work and was basing the performance off some anecdotal stuff. I will still double down on my models not being too high poly tho and that Jones is barking up an invisible tree and still has no idea how 3d models work.


My entire logic up until this point was that creating different lods would cause random frame drop popping and lag since that's what I've found from a small, quick and dirty test without making the lods properly. MXS doesn't have universal or auto downloads so that's why I haven't really cared about making lods for my models. Unless you download everybody's skins, it doesn't matter since you're only running your own skin. I had validation with Kellen's stream because he plays a full gate without frame issues. I will admit since it worked for him, I didn't pursue creating LODS since nobody runs skins like he does in the game. So if his game can run fine with a 40 rider gate with full skins and models, I guess I just assume I don't really need to sink any more hours into a model I make that is intended for personal use.

The "Kellen can't even load into the game" argument is nill until JLV fixes/updates his admittedly strong armed SCRAM file. As a content creator, I sell things at max resolution for personal use. As much as I tell people to not not run 4K files for a public release, i'ts not my fault if the still do. If you've ever made anything Jones you'd understand it's easier to downsize a file than to upsize it. I'm a third party in this and I'm sick of you saying I'm ruining the game to be honest. The files I make aren't exclusive to MXS, I make thing for this game but by a standard that other companies would be interested. I will take their word over yours. They are abiding by today's standards and have been of interest of the companies that you proudly claim are schooling me (an amateur 3d artist). Eat a bowl of dicks Jones, you still are straw manning another argument you don't have any part of. Try to find relevance somewhere else and stop stalking me you weirdo.

Alex, I never meant to imply you guys aren't super talented.

DJ, while you're here, I'm adding a new JS api that lets you pre-store poses so it doesn't have to continuously copy it over from Javascript. It's way faster, 400 waving flaggers go from 10 FPS to 90 FPS. Since I'm also adding interpolation, do you think the rotation matrix interpolation should:

A. Just interpolate the matrix element by element. Don't worry about shear and scale errors.
B. Square up and normalize the matrix after interpolating.
C. Square up, normalize and rescale so X, Y and Z scales are correctly interpolated.

I have A right now and it looks OK even though the interpolation would have introduced a little bit of unwanted scaling and shear. B and/or C wouldn't be hard to do but I'm not sure if it might accidentally erase an effect someone did intentionally. If some weirdo uses a shear matrix for some reason, A wouldn't break it like B and C would. What do you think?

jlv wrote:Alex, I never meant to imply you guys aren't super talented.

DJ, while you're here, I'm adding a new JS api that lets you pre-store poses so it doesn't have to continuously copy it over from Javascript. It's way faster, 400 waving flaggers go from 10 FPS to 90 FPS. Since I'm also adding interpolation, do you think the rotation matrix interpolation should:

A. Just interpolate the matrix element by element. Don't worry about shear and scale errors.
B. Square up and normalize the matrix after interpolating.
C. Square up, normalize and rescale so X, Y and Z scales are correctly interpolated.

I have A right now and it looks OK even though the interpolation would have introduced a little bit of unwanted scaling and shear. B and/or C wouldn't be hard to do but I'm not sure if it might accidentally erase an effect someone did intentionally. If some weirdo uses a shear matrix for some reason, A wouldn't break it like B and C would. What do you think?

Wow, that's a pretty big jump in performance. I can see why you were concerned with some of my other scripting suggestions.

When I did my interpolation function, I simply did it as per A, and its hard to visually pick up any distortion whilst its moving. I'm sure it depends on the object, and how much you're looking to change between keyframes. I think if you have large movements between keyframes, it's going to look a bit off anyway, so I think a simple interpolation would probably cut it. Can be easily fixed with more keyframes

Atleast I knew LOD’s work. But whatever makes you feel better dude. enjoy.

Quite messy, but since everyone throws in their opinion, I might just do the same.

For me this game really needs anti-aliasing and dynamic shadows. It would definitely help to simplify polygon meshes and still maintain smoother looks. Dynamic shadows are probably even more important, because as long as i can remember, we always had to keep things quite dark in order to create some sort of shadows. I mean we have lighting, but no shadows ...that's like suspension that only compresses, but doesn't come back out again.

I'm not 100% sure what it would take to implement more up to date rendering methods, and i don't want to offend anyone, but i can only speak for myself and guys like m121c and Jre when I say we lost interest simply due to a lack of graphic performance.
The first time I noticed how good my models could look, was when I upladed it to sketchfab. All of a sudden everything looked much better even though i didn't change a single thing. I'm sure you can make models look really good in sim (Jeremy and m121c have some fire materials), but the effort and amount of workarounds to get there is annoying, especially when you know there are simpler and more importantly much better ways available. I literally feel the urge to make new models, heck I even started on quite a few already but could never keep my motivation alive, because i could simply not look past the fact that i spend so much time on a model for a game, that doesn't give me back the graphics performance other games have from the outset (Motocross: Chasing the Dream).
I'm not trying to force anyone to develop the game further, after all I simply enjoy playing it, but if Sim doesn't get better graphics anytime soon, it will eventually make things worse than they already are, because as of now, all that keeps this game alive is it's superior physics, realistic racing and the opportunity to create content quite easily.
I don't follow everything like i used to, but to me it seems like Jeremy and aeffertz are the last creators who really provide HQ models, if they ever get tired of it ... well who else do we have?
I get that jlv tries to improve erode, animations etc. which is amazing and we can't thank him enough that he's still doing that, but I fear that without a solid graphics overhaul we'll lose the last talented people someday.

As far as too high poly models, one should tell apart actual detail and unnecessary "detail". I'd call myself quite old school when it comes to my modelling style, I learned from guys like DJ, ddmx, Wilson, Attacker and Jonathan, but I have to say when I look at models from Jeremy for example, they're as efficient as it gets! The poly count might have risen, but so did the quality and detail. Of course there are other creators who just turbo smooth the hell outta it, waste valuable polys on things you don't see (cables or such), while they keep the obvious things like plastics, frame, front end, etc. too low poly. I feel like the good modelers are getting shit even though they don't deserve it, while these other guys release shitty stuff and whenever someone points out that they're doing it wrong everyone protects them.
Take my word for it, Jeremy and aeffertz are doing an insanely good job, i mean just look how close they come to models made by actual professionals!

Andy_Hack wrote: I literally feel the urge to make new models, heck I even started on quite a few already but could never keep my motivation alive, because i could simply not look past the fact that i spend so much time on a model for a game, that doesn't give me back the graphics performance other games have from the outset (Motocross: Chasing the Dream).

I swear you got a second HDD with just half finish or started models

Mr. Wiggles wrote:

Andy_Hack wrote: I literally feel the urge to make new models, heck I even started on quite a few already but could never keep my motivation alive, because i could simply not look past the fact that i spend so much time on a model for a game, that doesn't give me back the graphics performance other games have from the outset (Motocross: Chasing the Dream).

I swear you got a second HDD with just half finish or started models

Shhh

Andy_Hack wrote:

Mr. Wiggles wrote:

Andy_Hack wrote: I literally feel the urge to make new models, heck I even started on quite a few already but could never keep my motivation alive, because i could simply not look past the fact that i spend so much time on a model for a game, that doesn't give me back the graphics performance other games have from the outset (Motocross: Chasing the Dream).

I swear you got a second HDD with just half finish or started models

Shhh

When you gonna come back on to discord and show me

Andy_Hack wrote:For me this game really needs anti-aliasing and dynamic shadows.

I do agree with the dynamic shadows, at least for the moving objects. I'd consider it a missing factor in the physics of the game, particularly for 3rd person players to get a judgement of the wheels proximity to the ground.

DJ99X wrote:I do agree with the dynamic shadows, at least for the moving objects. I'd consider it a missing factor in the physics of the game, particularly for 3rd person players to get a judgement of the wheels proximity to the ground.

x1000

Single most difficult thing about playing 3rd person is not knowing when your front wheel is slightly off the ground and immediately proceeding to whif it.