Support NeoGAF

[ZombieSupaStar]

Here's a demo that does dynamic branching, without the need for pixel shader 3.0, and still receives the huge performance boost that pixel shader 3.0 dynamic branching supposedly gives when utilized in a similar fashion.

Humus 3d Site

I dont know much about the technique but he says it basically can be done without PS3.0 and done with speed increases advertised in PS 3.0 also?


pana etc. any help on understanding this

 

[Chony]

Downloaded it. I can post pics if there is anywhere I can host it. Looks incredible and runs great.

 

[ZombieSupaStar]

took some myself

 

[Chony]

Sent, check your email, if you want more, just reply and ask for more.

 

[El Papa]

It looks like someone used the plastic wrap filter on those stones.

 

[ZombieSupaStar]

from chony

 

[Chony]

At first I couldnt figure out mouse aiming (I didnt click) as can be seen by those pics. After seeing the stones on the ground, there is no way that can be a flat texture, it looks too good.

 

[Mr. Lemming]

It's like a Tetris FPS.

Looks cool though.

 

[Mrbob]

N/M

 

[SKluck]

I can't believe my eyes!

Looks exactly the same with and without dynamic branching, obviously, because PS3.0 contains no new visual tweaks, afaik. But I go from 70 to 401 fps with it off/on. Which is fucking nuts.

 

[Smidget]

What are you guys running on that it is running this great? On my P4 3.0 gig with Geforce FX 5900 it's kinda crappy :/

 

[Teknopathetic]

<-- 9800 Pro, getting 2-300+ FPS with dynamic branching.

 

[TheQueen'sOwn]

<-- 9800 Pro, getting 2-300+ FPS with dynamic branching.

same... I get spikes though that exceed 700-800 .... it's weird o.0

 

[Mrbob]

Crytek seems to think dynamic branching is only feasable with PS 3.0....

 

[BeOnEdge]

no reason for glazed bricks. they are BRICKS.

 

[TheQueen'sOwn]

no reason for glazed bricks. they are BRICKS.

wet bricks? lol

 

[Chony]

I get 1200+ FPS when I make the window about the size of 150 X 150.

I am running a pretty old setup though:
AMD Athlon 1 ghz
Radeon 9500 non pro soft modded to Radeon 9800
512 mb RAM

I think this demo is pretty much entirely graphics card dependant, so if you dont have a good graphics card, you will get very poor performance.

 

[ImNotLikeThem]

see, the only problem i have with things like this, besides the super shinyness (but thats another story) is the edges of polygons. they need to find a way so that the "bumped" effect actually does add volume to something, so that the edges of walls and such dont appear to be a jagged flat edge while the inside of the wall/floor has depth and volume and such. thats my only complaint, because it takes you out of the scene when you see that.

 

[Teknopathetic]

It should be noted that dynamic branching isn't the actual graphics itself, the only purpose of this tech demo is to show off Dynamic Branching, which it did. I'm sure devs would use bump/normal/etc. mapping to a more believable degree in a real game.

 

[HyperionX]

Know how to use tabs in Firefox?

 

[Hiro_Kunimi_80]

it's perfect for boobs and Doa4

 

[Bregor]

I saw this a couple of weeks ago. It really makes me wonder if PS 3.0 is going to have any significant effect on PC games at all.

 

[Panajev2001a]

Humus
Senior Member
Senior Member


Joined: 05 Feb 2002
Posts: 2607
Location: Toronto, Canada

PostPosted: Thu Jul 01, 2004 9:45 pm Post subject: Reply with quote
Thowllly wrote:
I don't think he is claiming that SM2.0 will look and perform similarly to SM3.0, just that it can look and perform similarly.


Right. Not always, but in many situations.

Dio
Senior Member
Senior Member


Joined: 01 Jul 2002
Posts: 1368
Location: UK

PostPosted: Thu Jul 01, 2004 10:12 pm Post subject: Reply with quote
Zeno wrote:
It is a special case which requires switching a bunch of render states and sending all the geometry through the pipe again.

Remember that there is no point in optimising if the optimisation doesn't help you.

For example, if you're running a 200-instruction pixel shader, there's no point jumping through hoops to avoid using floating point buffers, because the pixel shader takes so long to execute that the extra bandwidth taken is irrelevant.

Similarly, if you are limited by your pixel shader, then it's a lot less likely to matter if you send all the geometry through many times (if they don't share resources).

In contrast, if your average polygon size is 2-3 pixels, then it doesn't really matter if you run a hundred instruction pixel shader on it. It won't cost anything.

It's all about pipeline balances.

Humus is basically putting more workload on the Vertex Shader to use the static branching capabilities ( static = you branch based on a value you receive as input and that you cannot modify inside the Shader ) of the GPU to optimize the pixel shader.

As long as you are limited by the Pixel Shader this is a good optimization.

The major features of Shader Model 3.0 are Geometry instancing and Vertex Texturing (which is impossible on Shader Model 2.0 cards as the Vertex Shaders do not have any associated TMU ).

 

[6.8]

Impressive, but it has bad art.

 

[bbyybb]

Well it is more of a tech demo, and they dont have to look pretty to test your hardware.


Cheers,
bbyybb.

 

[tedtropy]

Know how to use tabs in Firefox?

Tabbed browsing is overrated. I like to be able to Alt+Tab through my pages.

 

[Scrow]

Tabbed browsing is overrated. I like to be able to Alt+Tab through my pages.

errr... how is ctrl+tabbing through your tabs in firefox any different?

 

[Mason]

Oh, I love this guy's website. He has some great tech demos. I wish he'd work on an open-source game engine or something.

 

[epmode]

errr... how is ctrl+tabbing through your tabs in firefox any different?

well, you lose that quarter to half-inch of space that displays the tabbed page titles. i guess.

 

[Sumasshu]

I use the side buttons on my mouse to tab through pages. *I am special*