Evidence of Valve using LODs again!

[Rexy]

Evidence of Valve using LODs again in TF2!

It's true! They're using LODs again in TF2!! Take a look!

stone_window001:

And also variants stone_window001a and b have this same LOD use.

But that's it. None of the other assets in this update have LODs. :blushing:

Edit:
I've removed the sections about the poly count for the resupply cabinet, while still high, I'm to blame for blowing the reality of the poly count out of proportions, due to my pure ignorance. Apologies, all.

 

[Ravidge]

:U

 

[Icarus]

:V

 

[zpqrei]

:W

 

[Dr Forrester]

:Q

 

[Mr.Blob]

 

[Tekku]

:|

 

[DaBeatzProject]

:

 

[Khuntza]

:X

 

[Engineer]

:{O

 

[J4CK8]

They look like new crafting blueprints

 

[Ida]

It's almost like Valve care more about churning out new stuff all the time just because they feel that they need to, rather than spending time to ensure that their game's quality remains consistent. It's kinda sad.

But at least one modeler has done a proper job, I guess?

 

[grazr]

Havn't we already discussed how polygons are fairly cheap to render, it's the materials that are expensive.

We're nearing the DX10-11 generation where many people are running 64 bit systems. That coupled with the fact we'll possibly see an engine upgrade with Ep 3...

 

[Ravidge]

Havn't we already discussed how polygons are fairly cheap to render, it's the materials that are expensive.

We're nearing the DX10-11 generation where many people are running 64 bit systems. That coupled with the fact we'll possibly see an engine upgrade with Ep 3...

Polygons are cheap to render, but it's definitely not without cost at all.
As I tested before with these things http://dl.dropbox.com/u/1281220/SS-2010.12.07-20.51.38.jpg too many polygons really cripples the framerate. These polyballs have a texture of 8x8 pixels and no normalmap or reflective properties. So the texture in this example is extremely cheap. much more so than the meshes.
Textures at least have automatic mipmapping optimization.

It's a unrealistic example though, but LODs are still incredibly helpful.
Ignoring optimization just because computers are getting more powerful is a pretty lame excuse. My system was almost brand new when TF2 was released and I could run the game smooth as silk, but it's getting worse for every year and it shouldn't. I can't even choose to use low quality models now, because there are no high-medium-low options anymore, everyone gets the highpoly model no matter what.

 

[grazr]

Oh no, i totally agree that we need LOD's. I was simply saying that perhaps we'll see some performance improvements when and if the engine gets improved with ep3. After all, the engine has its own limitations as does your computer hardware. Try running a HL2 scene (just geoemtry) in HL1 and you'll find the performance horrible in HL1 despite the fact that the scene (obviously) costs the same amount of memory.

 

[HellJumper]

c-c-c-c combo breaker!

On a more serious note.
I really love the math going on in this thread.

 

[Fruity Snacks]

The poly count is high.... i think...

BUT...

Whats LODs?

Edited because I sounded noobish. (which I am)

 

[grazr]

LoD = level of detail. At greater distances there is no need for high levels of detail so there are several versions of a single prop; at X distances the level of detail of said prop is reduced (to a version with fewer polygons) to increase performance.

The concept is not limited to models, there are solid entities that incorporate similar resource saving features, such as func_lod which allows you to assign custom fade distances to solid geometry. Although that's not physically the same process (of saving tri's) the affect is the same: better performance through reducing detail at long distances.

 

[EArkham]

The thing I'm most disappointed with by all this LOD business is that I missed out on the combo on the first page.

:ひ

Kep

 

[Fruity Snacks]

Ah, alright, thanks grazr, that makes sense.

@ Turban smiley!