shader competition results

[Simon F]

People may not even realize they're in point sampling mode. You're overestimating how much knowledge people have.

We're not talking about "people." We're talking about people whose job is programming, specifically programming in 3D. We're talking about people whose job it is to know how to properly program in 3D.

And have you always written 100% bug-free code? When under pressure of deadlines from the evil publisher who's breathing down your neck? :|

To paraphrase an old cliche, "You can't eliminate all of the bugs all of the time" :|

Furthermore, I don't think everyone in the industry could possibly know all the "tricks of the trade" especially when there are more and more new tricks each year! Someone is sure to get some wrong <sigh>

 

[KimB]

Okay, tell me, how obvious would it be if point sampling was used by accident?

Seriously, how obvious?

This is not a bug that should make it through development. Point sampling will pretty much certainly only be used when it is necessary to produce the desired result.

 

[Simon F]

Okay, tell me, how obvious would it be if point sampling was used by accident?
Seriously, how obvious?

Because it might have been deliberately set for one small effect and then accidently left on for other objects.

 

[KimB]

But how obvious would it be that point sampling was enabled by mistake? This sort of bug would be noticed in a heartbeat. I say your claim that it could actually happen by mistake requires evidence. I have yet to see any game that accidentally used point sampling.

But there are a number of games and demos that require point sampling, and have problems with anisotropic is forced over point sampling.

It's a simple numbers game. Forcing anisotropic filtering over point sampling is bad, because point sampling is used where it's useful. Currently. No mights, maybes, ifs about it.

 

[Simon F]

But how obvious would it be that point sampling was enabled by mistake? This sort of bug would be noticed in a heartbeat. I say your claim that it could actually happen by mistake requires evidence. I have yet to see any game that accidentally used point sampling.

If you insist, I personally feel that the point filtering used on "The house of the dead" was a mistake however I was told that it "was an artistic decision". I've seen other examples in games where I've thought the same thing.

But there are a number of games and demos that require point sampling, and have problems with anisotropic is forced over point sampling.

I'm not disputing that, however, I'm saying the converse also exists.

 

[Joe DeFuria]

We're not talking about "people." We're talking about people whose job is programming, specifically programming in 3D. We're talking about people whose job it is to know how to properly program in 3D.

Surely.

Then how come the vast majority of these "people" don't put in native support for AA and Anisotropic filtering then?

Just because "it's your job", doesn't mean you do things right all the time...

 

[Joe DeFuria]

But there are a number of games and demos that require point sampling, and have problems with anisotropic is forced over point sampling.

Then turn off forcing in the driver control panel. Problem solved.

 

[flipper]

back on track?

not to re-direct the current topic in the thread, but i too would just like to say congrats to all of the winners in the compo. my fav was also frogger, just because it was such an unusual idea to attempt.

btw, i also had the blinking vertical lines problem mentioned in the news post thread. i took a quick look into the shader file and found out that the code causing the problem was an if statement used to draw the time bar in the Display shader. when i replaced this code:

if( min(Data.y * 0.003f - BarPos.x, min(BarPos.y, BarPos.x)) >= 0 )
Colour = float4(1,0,1,1); // purple

with this code:

if( In.ScreenPos.y > 0.73f &&
In.ScreenPos.x > 0.65f &&
(In.ScreenPos.x - 0.65f) < (Data.y * 0.003f) )
Colour = float4(1,0,1,1); // purple

everything displayed properly on my cat 3.8, 9700 PRO setup.

 

[Humus]

Surely.

Then how come the vast majority of these "people" don't put in native support for AA and Anisotropic filtering then?

Just because "it's your job", doesn't mean you do things right all the time...

The problem is this:
First no games were written with AF in mind because AF wasn't supported. Then once hardware support appeared the IHVs of course wanted it to be used, but since no games supported it they decided to add a force option in the driver. Enough time passed for the idea that AF is a thing you set in the driver's control to stick. Now when developers write games, putting AF options in the game isn't high priority since it's already supported through the driver's control panel. It sux, but it's the way it is.

 

[KimB]

If you insist, I personally feel that the point filtering used on "The house of the dead" was a mistake however I was told that it "was an artistic decision". I've seen other examples in games where I've thought the same thing.

Point sampling is still just way too obvious to just overlook. There's no way it was a bug. There was just somebody on the design team who had a really wierd idea of what looked good (played too many Playstation games, perhaps?). Anyway, for you to convince me, you're going to need to find a PC game that shipped using point sampling, and later the game was patched to support bilinera filtering.

I don't think any exist.

Hell, since before Direct3D and OpenGL were used in any PC 3D games, bilinear filtering had no performance hit. Bilinear filtering was the major visual difference between software rendering and hardware. Sure, there was also the higher resolution, but the real difference was the filtering. I don't think that mistake has been made (well, made and not noticed, at any rate).

 

[Joe DeFuria]

Now when developers write games, putting AF options in the game isn't high priority since it's already supported through the driver's control panel. It sux, but it's the way it is.

Certainly agreed. Although it's not already supported exactly how everyone wants it to be. It was exactly offered as a way to use it for legacy games where devlopers had no reason to implement it because hardware at the time didn't support it.

Now, if developers continue to "rely on the control panel" for AA / ansio, even though just about every modern GPU has supported on form or another for the past few years, then they (software devleopers) deserve for unexpected things to happen. It sucks, but that's the way it is.

 

[DeanoC]

Hell, since before Direct3D and OpenGL were used in any PC 3D games, bilinear filtering had no performance hit.

Nope I shipped a D3D5 game where we still had the option to turn bilinear filtering off, as it was killed framerate on ATI Rage 2 chipsets...

Was a few years ago mind....

 

[KimB]

Well, yes, I suppose at that time there were some "3D decelerators" like ATI's Rage, and S3's Virge.

And ATI's Rage chipset had such high saturation due to massive amounts of OEM deals...well, let me just say, thank God for nVidia.

 

[Althornin]

well, let me just say, thank God for nVidia.

 

[ET]

Well, yes, I suppose at that time there were some "3D decelerators" like ATI's Rage, and S3's Virge.

And ATI's Rage chipset had such high saturation due to massive amounts of OEM deals...well, let me just say, thank God for nVidia.

NVIDIA didn't have anything on the market in the days of the Rage II+. And, BTW, NVIDIA did prove with the Rive128 that the most important thing for gamers was speed, and not image quality, and they've learned that lesson well. It's still true today.

Not that I'm saying that ATI had great image quality -- they had that awful lack of bilinear for alpha. A real pity since the Rage Pro was such an impressive chip. If not for that and the awful drivers, it likely would have been better regarded. At least ATI learned from that.

 

[ET]

People may not even realize they're in point sampling mode. You're overestimating how much knowledge people have.

I assume that people who force anisotropic are people who can tell point sampling from bilinear. If they can't, there's no point anyway in forcing it. So either way overriding point sampling would be meaningless.

I use Radlinker.
seperate link for each game.
Those games that have aniso/AA controls, i dont need to use the forced settings.

So I guess you didn't have any problem with Frogger, since you didn't turn aniso on by default. But you might agree that most people who use just the driver option for AF will have it forced on for everything by default.

Now, if developers continue to "rely on the control panel" for AA / ansio, even though just about every modern GPU has supported on form or another for the past few years, then they (software devleopers) deserve for unexpected things to happen. It sucks, but that's the way it is.

It's the users that suffer from the issue, not the developers.


IMO the default that is most beneficial should be used. If changing point sampling to anisotropic causes problems for more applications than it solves, then it should be off by default. It should be a separate option, in any case. That's because it's obviously harmful to some applications, and people would like the option of having AF for textures that need it, and not having it for those where it causes harm.

 

[KimB]

NVIDIA didn't have anything on the market in the days of the Rage II+. And, BTW, NVIDIA did prove with the Rive128 that the most important thing for gamers was speed, and not image quality, and they've learned that lesson well. It's still true today.

Um, then why did nVidia dramatically improve image quality with the TNT?

Regardless, what I meant was that nVidia drastically raised the bar for performance and image quality in the low-end market, starting with the TNT2 M64, and much moreso with the GeForce2 MX. Subsequent products have cemented this commitment to high-performance, high-featured low-end products (for the price).

 

[ET]

Regardless, what I meant was that nVidia drastically raised the bar for performance and image quality in the low-end market, starting with the TNT2 M64, and much moreso with the GeForce2 MX. Subsequent products have cemented this commitment to high-performance, high-featured low-end products (for the price).

Conveniently forgot the 4mx?

Well, it was a jab, and doesn't mean that NVIDIA always does it (certainly with the GeForceFX 5200 they went for features over speed). However, I think it's quite correct that most people care more about speed than image quality. People here excepted, of course.

 

[KimB]

Regardless, what I meant was that nVidia drastically raised the bar for performance and image quality in the low-end market, starting with the TNT2 M64, and much moreso with the GeForce2 MX. Subsequent products have cemented this commitment to high-performance, high-featured low-end products (for the price).

Conveniently forgot the 4mx?

No. That doesn't change the fact that nVidia started the trend.

And the disappointment of the GeForce4 MX was only due to comparison to other nVidia products. Nobody else had, at that time, as good of low-end parts.

 

[Joe DeFuria]

Now, if developers continue to "rely on the control panel" for AA / ansio, even though just about every modern GPU has supported on form or another for the past few years, then they (software devleopers) deserve for unexpected things to happen. It sucks, but that's the way it is.

It's the users that suffer from the issue, not the developers.

No, because any user who switches their control panel from the default "application preference" surely knows how to switch it back.

Users only suffer if the developers are lazy.