Tim's thoughts

Hyp-X said:
Halo runs very good on a Geforce 3 with a 733MHz Intel Celeron processor ;)
People will always take for granted that XBox is just a PC in a box, when it's actually very far from that.
Well, it's closer to a GeForce4 with low memory bandwidth. And a large part of the performance comes from the simple fact that the programmers get much lower-level access to the hardware.
 
Personally I think more extensive shaders are only icing on the cake anyway ... the amount of developers which use high poly modeling with automated generation of bumpmaps for per pixel lighting is still pretty damn small. Most developers arent being slowed down by slow shaders IMO.
 
MfA said:
Personally I think more extensive shaders are only icing on the cake anyway ... the amount of developers which use high poly modeling with automated generation of bumpmaps for per pixel lighting is still pretty damn small. Most developers arent being slowed down by slow shaders IMO.
Polycounts are much harder to scale over different hardware than pixel shader effects. So, polycounts are currently limited by the slowest hardware the game supports. This will continue to be the case until robust HOS support is available.

So, it is more realistic to expect more advanced pixel shading to take the place of increasing polycounts significantly for a while yet.
 
"Most developers arent being slowed down by slow shaders IMO."
------------------------------------------------------------------------------------
but this doesn't mean they aren't slow, and they certainly aren't overly complex.


"From what I heard porting Halo to the PC was hell.
There's no surprise the hardware requirement went way up for doing the same amount of work."
-------------------------------------------------------------------------------------
but they aren't doing the same amount of work. haloPC has ps2.0 while haloXB only has 1.1. much of the rendering engine was redone, so it's not an apple to apples comparison.

i guess i shouldn't have used halo as an example. with any game there are people that are satisfied and dissatisfied with it's performance because it is subjective and partly related to the gaming experience. here's a more synthetic example. check out humus' water demo...

http://esprit.campus.luth.se/~humus/
"Required Direct3D capabilities:

Vertex shader 1.1
Pixel shader 2.0
RG16F render targets"

on my r9700 i can get over 2000fps while the water plane is not on screen, but when the water is covering all of the screen i get about 60-65. from the default camera angle (water covers aprox 50% of screen) the framerate jumps around but looks to average about 130-150. there are no shaders on the skybox where i'm getting 2000fps, the performance hit for this single plane with a ps2.0 effect on it is insane! it's 30 times slower with the shader covering the full screen.

this effect is the best interactive pixel based water effect i've seen. but as nice the demo looks, it's still not what i'd consider a "extremely complex pixel shader". it's just pushing the limits of what this generation of hardware can do.
c:
 
True, but such predictions must be made by game engine developers. In this sense, Tim Sweeney didn't do as good of a job predicting the direction of hardware development as John Carmack did.

I don't think I want to go there. What I can say is that I'm looking forward for CroTeam's Serious Sam 2 engine next year ;)
 
Chalnoth said:
MfA said:
Personally I think more extensive shaders are only icing on the cake anyway ... the amount of developers which use high poly modeling with automated generation of bumpmaps for per pixel lighting is still pretty damn small. Most developers arent being slowed down by slow shaders IMO.
Polycounts are much harder to scale over different hardware than pixel shader effects. So, polycounts are currently limited by the slowest hardware the game supports. This will continue to be the case until robust HOS support is available.

So, it is more realistic to expect more advanced pixel shading to take the place of increasing polycounts significantly for a while yet.

I'm afraid that all predictions so far considering robust HOS support were wrong and the thought that even the next two generations of cards will not deliver what was initially planned scares me. IHVs are definitely in a tight spot fitting it all into each cases die restrictions. What one would question here are priorities and what the majority of the developers is actually asking for.

It seems weird at best that if there would be repeated requests from developers for advanced HOS and IHVs would just ignore them or set their priorities differently...
 
You can dress low poly models up with as many gimmicky effects as you want, but the point where lack of polys and more importantly good normals drowns out potential improvements by shader tricks is reached pretty quickly IMO.

At some point you have to get the basics right before you can move on ... and I think that point has been reached. They dont need DX9 shaders till they make decent use of what DX8 made possible as far as per pixel lighting is concerned IMO.
 
"What I can say is that I'm looking forward for CroTeam's Serious Sam 2 engine next year"
------------------------------------------------------------------------------------
ns. i can say about next year what i said earlier this year about this year. (can i fit another year into that sentance?). it's going to be an interesting year. DooM3, Half-Life 2, Serious Sam 2, and most likely a new lithtech engine.
c:
 
I have never seen a pc game that was geometry limited on a modern graphics card. On consoles, sure, but on the PC? Most modern games would probably barely tax the vertex shader of the original GeForce3. I really have no idea why pc-developers like low poly environments so much, maybe it's because they like to make lots of Direct3D calls and become cpu-limited quickly because of that?

In contrast, the GeForce 2MX which sold in droves was basically always memory bandwidth limited even at really low res and most of the budget cards that sell well today can pump out a decent amount of triangles but have paltry fillrate and/or bandwidth. I honestly suspect the main reasons for having fewer triangles is content creation time. Creating detailed geometry takes time.
 
GameCat said:
I have never seen a pc game that was geometry limited on a modern graphics card. On consoles, sure, but on the PC? Most modern games would probably barely tax the vertex shader of the original GeForce3. I really have no idea why pc-developers like low poly environments so much, maybe it's because they like to make lots of Direct3D calls and become cpu-limited quickly because of that?
The reason is exactly what I explained before. Polycount is limited by the lowest common denominator hardware for which the game was designed. If we had robust higher-order surfaces support, polycounts could adapt to the hardware, and so even the most powerful hardware could be significantly taxed.
 
MfA said:
You can dress low poly models up with as many gimmicky effects as you want, but the point where lack of polys and more importantly good normals drowns out potential improvements by shader tricks is reached pretty quickly IMO.
Well, the question remains as to what you consider low-poly. If you look back to nVidia's Werewolf demo, the wolf really didn't have that many polygons, not nearly as many as you'd think from first glance. Shader tricks can help quite a lot as long as there's not too few polygons.
 
MfA said:
Not everything has fur to cover rough edges.

Yes and not everyone wants every game to be composed of nothing more than four badly animated, but pixel perfect characters on screen at a time.

If anything, Shading should compliment increasing polygon counts, not be a substitute as it has almost become.
 
Diminishing returns are a significant factor with poly counts. Pretty soon games will get to a point where polygon counts are "good enough" (I define this being when games use charachter models equivilant to the Dawn demo), but we won't reach a point where lighting is "good enough" for a long time...
 
nobie said:
Diminishing returns are a significant factor with poly counts. Pretty soon games will get to a point where polygon counts are "good enough" (I define this being when games use charachter models equivilant to the Dawn demo), but we won't reach a point where lighting is "good enough" for a long time...

Perhaps. Yet untill I can run into a crowd of thousands in a GTA like game - I'm inclined to call Bullshit!. I think the PC, perhaps strongly due to it's architecture, has traveled down a path where instead of creating geometry heavy worlds that are filled with many objects; has been more inclined to have few onscreen objects, which are highly detailed.

Your comments are only supporting this ideology. Which, IMHO is almost ignorant when you think of what could be done. And when you think of the 'sub-genre' of games which are becoming popular (GTA like, open ended worlds) the limitations are only more appearent. Maybe one Dawn is pretty and cool to talk about, but make me a city out of them and then gaming will have progressed,
 
Right now you can have one Dawn in real time on the fastest PC's with the fastest gfx cards.
Tomorrow it will be two..
Buttt........ whatabout the background, AI, physics?
Dawn is a demo and even she doesnt look all that realistic when you compare her to a photograph.
Pretty soon is normally translated to "5 years time" ;)

Oh and Sweeney descriped kind of happened on PS2, and may happen on PS3 but PC Gfx Cards didn't go in that direction.
 
Vince said:
Yes and not everyone wants every game to be composed of nothing more than four badly animated, but pixel perfect characters on screen at a time.
How well the animation is done doesn't have much at all to do with polycounts.
 
Chalnoth said:
Vince said:
Yes and not everyone wants every game to be composed of nothing more than four badly animated, but pixel perfect characters on screen at a time.
How well the animation is done doesn't have much at all to do with polycounts.

Think about facial animation, think about Silent Hill 3 ;)
 
Panajev2001a said:
Chalnoth said:
Vince said:
Yes and not everyone wants every game to be composed of nothing more than four badly animated, but pixel perfect characters on screen at a time.
How well the animation is done doesn't have much at all to do with polycounts.
Think about facial animation, think about Silent Hill 3 ;)
Intricate facial animation isn't something you're going to need on every character on the screen all the time, so if that's one of the things that is important for the game, it can be done well without a huge overall polycount.
 
Back
Top