WIP: Bradley Fighting Vehicle

[Eghi]

I put some screenshots in this thread:
https://www.benchmarksims.org/forum/showthread.php?22653-WIP-Leopard-2-A5&p=330331&viewfull=1#post330331
I think should make a new thread for this vehicle and get to know your opinions.
What to add, what is not needed?? etc …

This is M2:

LOD1 - 1106 tris
LOD2 - 851 tris
LOD3 - 315 tris
LOD4 - is a very basic untextured simple representation of the overall shape.
M3 (WIP):

Slightly more tris but still below 2k

Destroyed models WIP.

Cheers,
EGHI

[neystratiou]

Jesus Arek,
Do you want to kill all of us? :rofl:
You are a tank “buster”… Great work again mate!
Nikos.

[PimpDaddyStalker]

Wow, almost too pretty to strafe.

Fine work.

[Snowman]

Nice!!

Looks very good!

Thx

[lazystone]

What should I say?

… high quality stuff, Arek !

Once again, T(h)anks a lot … lot …lot.

Cheers,
LS

[Dee-Jay]

Exactly what BMS needs … some VERY GOOD quality low poly models.

[BANITA]

wow respect,you good

[Eghi]

Test in Falcon:

I wonder improve the wheels, or leave it?

[Dracfalcon]

I am no expert…but looks good to me.

[Shadow]

Looks Great to Me

[Raptor]

Zoom into the vehicle through TGP, if you see something strange then will sure benefit from some additional tris.

[lazystone]

@Eghi:

I wonder improve the wheels, or leave it?

Good to go, IMHO.

… but +1 to what Raptor said, … to make sure.

Cheers,
LS

[l3crusader]

@Raptor:

Zoom into the vehicle through TGP, if you see something strange then will sure benefit from some additional tris.

+1, but I’m pretty sure these are already just fine

[Eghi]

Guys,
Thanks for valuable advice. The wheels is very low tris ATM. I had in mind more details (more triangles). As Cruz said, should be fine … But indeed worth doing all the tests…

Cheers,
EGHI

[Raptor]

Just saying, what if the wheels where nicely painted in the skin, while the external “part” of the wheel skin is transparent, so the actual external wheel triangles will be invisible…

[lazystone]

IMHO bad idea, Raptor.

… because it will cost more performance to use the alpha channel
than to add a few polys (if really needed).

Technically speaking from what I’ve learned a few weeks ago,
and how I understood, …

Using alpha channel here will add another PType,
and every PType change (beside other things) will force a new drawcall
to the DX engine.

It’s not the number of polys/tris, it’s the number of drawcalls
per model which affects the performance.

I hope the BMS core modellers will correct me if I’m wrong here.

Cheers,
LS

[Raptor]

The idea was (again) from EMF devs almost 7 years ago, and as you can see it works well. Maybe I am not the best 3D modeler to describe the small details, but 1 pic = 1000 words.

Something similar is implemented to some real simulators too, like the Block 52+ Advanced I’ve flown recently. If you try to fly below the huge Rio bridge you will explode, because in order to save some extra graphical power they build the model exactly this way, a solid model with low poly’s while a detailed and transparent skin so to see what’s behind the structure. Bottom line, you’ll not notice any differences.

[lazystone]

@Eghi:

Guys,
Thanks for valuable advice. The wheels is very low tris ATM. I had in mind more details (more triangles). As Cruz said, should be fine … But indeed worth doing all the tests…

Cheers,
EGHI

Hehehe, Arek, I can feel your pain.

Let’s say you use a 8 segmented cylinder intead of the actually 6 segmented one,
then you’ll have nearly 100 more tris.

So if you’re really not happy with the 6 segmented one,
just do it, as long as you’re within BMS specs.

Cheers,
LS

[lazystone]

@ Raptor

… maybe BMS is working differently? … I don’t know.

But I fully understand your example, and in past I did many models the same way.

Today I’m not sure that if that bridge would have been modelled
with geometry only (no alpha) would perform better.

Assuming the bridge has 120 support cables, and I can create one cable
with a 3 segmented cylinder (no caps), then 120 x 6 (2 tris per side) = 720 tris.

I think 720 more tris are better than using alpha, but honestly in this case
I would ask the BMS modelling gurus.

Cheers,
LS

[WaveyDave]

@lazystone:

Using alpha channel here will add another PType,
and every PType change (beside other things) will force a new drawcall
to the DX engine.

It’s not the number of polys/tris, it’s the number of drawcalls
per model which affects the performance.

I hope the BMS core modellers will correct me if I’m wrong here.

I will gladly point out that a break in the batch (and subsequent drawcall) is caused by many things not just a change of pType

In order for triangles to end up in the same batch, they must have

• same world matrix
• same view matrix
• same texture
• same material
• same shader
• same ptype (vertex type)
• same renderstate
• same smooth group

Regards
Dave