r420 may beat nv40 in doom3 with anti-aliasing

[radar1200gs]

_PP is a big deal because when DX9 was devoloped it was envisaged that any shaders not requiring full precision (I.E: everything devs currently use PS1.x for in their current "DX9" games) would run through it and maintain high quality & good precision rather than drop back to DX8 shaders (which aren't even part of DX9 in the first place).

 

[Dave Baumann]

_PP is a big deal because when DX9 was devoloped it was envisaged that any shaders not requiring full precision (I.E: everything devs currently use PS1.x for in their current "DX9" games) would run through it and maintain high quality & good precision rather than drop back to DX8 shaders

FP16 actually has less precision than some integer modes offered out there.

(which aren't even part of DX9 in the first place).

Its a requirement of DirectX to have support for all previous shader models (whether thats through superset or specific support is up to the IHV to decide).

 

[digitalwanderer]

_PP is a big deal because when DX9 was devoloped it was envisaged that any shaders not requiring full precision (I.E: everything devs currently use PS1.x for in their current "DX9" games) would run through it and maintain high quality & good precision rather than drop back to DX8 shaders (which aren't even part of DX9 in the first place).

****ERRRRRRRRRRRRRRRRRR!*****

Wrong answer, although that one sounds a lot better than the truth.

_PP was added late to the dx9 spec to save nVidia's ass after their failed attempt to re-write the industry standards, any other interpretation of it is simply trying to re-write history Radar. 8)

 

[radar1200gs]

Doesn't matter if it was early or late. Fact is its part of DX9 and has been since DX9 was publiclly released.

Of course DX9 is backward compatible (as per usual) with previous versions but that doesn't mean you can use DX8 features and say you have a DX9 app.

 

[digitalwanderer]

I don't think nV35 or any of the nV3x generation actually are dx9 cards. They may SAY they are, and they might be physically capable of all the dx9 things....but they can't do it at any kind of playable framerate or with any kind of decent AA.

Now that's just my opinion, but it is an opinion I based on my own experiences with my own nV35.

Sorry, I just liked being able to finally say/read that. It was worth the price. 8)

 

[Dave Baumann]

Of course DX9 is backward compatible (as per usual) with previous versions but that doesn't mean you can use DX8 features and say you have a DX9 app.

Exactly what planet are you on? If you have a requirement for for DX9 then you need DX9 ergo its a DX9 application. Every game out there still uses texturing as its foundation (and most will run without shaders at all) - by your logic games can only claim to be DX5/7!!

 

[radar1200gs]

Thats true enough and a big part of why the TNT-2 M64 and GF2MX live on.

 

[radar1200gs]

You can't use a PS1.x effect in a game and claim it's a DX9 effect, even if there are other SM2.0 shaders in the game.

Your game is a mixed mode game in that case. To be a pure DX9 game it needs to use only SM2.0 shaders.

 

[digitalwanderer]

You can't use a PS1.x effect in a game and claim it's a DX9 effect, even if there are other SM2.0 shaders in the game.

Your game is a mixed mode game in that case. To be a pure DX9 game it needs to use only SM2.0 shaders.

Where exactly did you pull that definition from? (I'm guessing your butt, but I figure I'd be nice and ask. )

Are you defining the industry terms now? Does that mean there are no dx9 games out?

 

[radar1200gs]

There are some games that make limited use of DX9 features, but no pure DX9 games.

It shouldn't be difficult for an ATi supporter to modify a game to be pure DX9 though, just replace all 1.x shaders with 2.0 shaders.

I'm sure your hardware will cope, the ATi supporters are fond enough of telling us how good their DX9 support is.

Of course, this begs the question, why do devs use PS1.4 in R3xx paths if the above is true?

EDIT: Might I suggest the much beloved by ATi supporters 3DMark03 as a starting point for this exercise

 

[LeStoffer]

There are some games that make limited use of DX9 features, but no pure DX9 games.

It shouldn't be difficult for an ATi supporter to modify a game to be pure DX9 though, just replace all 1.x shaders with 2.0 shaders.

That doesn't much sense to me. I think you might mistake a DX9 level game engine for being one that only renders with pixel shaders. No true. DX is a frameset of specs and features for input control, audio, display and what not. Shaders are only a small part of it all and simple old school (multi) texturing is not dead or gone at all.

You could in theory have one single shader running at PS 2.0 and the rest of the rendering being based on simple DX6 multi texturing and you would then need DX9-hardware to run it. Got it?

 

[FUDie]

I beg to differ about _PP not being a major DX9 feature.

If all DX9 graphics cards supported _PP, you would see more titles using SM2.0 and less using PS V1.4. Like I have said before, the lack of _PP on all cards has held back the adoption of DX9.

Your logic is flawed (what's new?). The R3x0 based cards support _PP. Do shaders with _PP modifiers suddenly cease to work on R3x0 cards? No. Since the R3x0 cards were faster running 24-bit precision than the NV3x was running _PP, your "argument" doesn't hold water (as usual).

-FUDie

 

[digitalwanderer]

There are some games that make limited use of DX9 features, but no pure DX9 games.

It shouldn't be difficult for an ATi supporter to modify a game to be pure DX9 though, just replace all 1.x shaders with 2.0 shaders.

I'm sure your hardware will cope, the ATi supporters are fond enough of telling us how good their DX9 support is.

Of course, this begs the question, why do devs use PS1.4 in R3xx paths if the above is true?

EDIT: Might I suggest the much beloved by ATi supporters 3DMark03 as a starting point for this exercise

You do realize that what you posted makes no sense whatsoever, right? :|

 

[FUDie]

There are some games that make limited use of DX9 features, but no pure DX9 games.

It shouldn't be difficult for an ATi supporter to modify a game to be pure DX9 though, just replace all 1.x shaders with 2.0 shaders.

I'm sure your hardware will cope, the ATi supporters are fond enough of telling us how good their DX9 support is.

Of course, this begs the question, why do devs use PS1.4 in R3xx paths if the above is true?

EDIT: Might I suggest the much beloved by ATi supporters 3DMark03 as a starting point for this exercise

Some people never learn, do they?

From my experience on R3x0 based cards, PS 1.4 shaders run at the same speed as the equivalent PS 2.0 shaders. I say equivalent because some PS 1.4 instructions don't map exactly to PS 2.0 (input modifiers for example).

You "forget" that the R3x0 chips are running FP24 all the time. Hence, it makes no difference whether the source shader is PS 1.4 or PS 2.0.

You consistently make lots of "claims" like that in your last paragraph and provide no support whatsoever. I'd call that trolling.

-FUDie

 

[DegustatoR]

Your logic is flawed (what's new?). The R3x0 based cards support _PP. Do shaders with _PP modifiers suddenly cease to work on R3x0 cards? No. Since the R3x0 cards were faster running 24-bit precision than the NV3x was running _PP, your "argument" doesn't hold water (as usual).

-FUDie

Yeah, right. And all FX cards do support anisotropic level 16 then: if you set 16 in the API you'll still get anisotropic!

R3x0s DO NOT support _pp hints. They just IGNORE them. That is a big difference.

 

[FUDie]

You can't use a PS1.x effect in a game and claim it's a DX9 effect, even if there are other SM2.0 shaders in the game.

Your game is a mixed mode game in that case. To be a pure DX9 game it needs to use only SM2.0 shaders.

So if a game uses PS 2.0 for everything except the HUD it's still a "mixed-mode" game? You should be happy that games are "mixed-mode" because they are what has kept NVIDIA in the game for the last year! Suppose Gun Metal were a real DX9 game? Suppose Far Cry didn't have a fallback path for NV3x parts. Suppose Tomb Raider had a benchmark mode (oh wait, it did, didn't it? ). NVIDIA is the reason for the fallbacks, not ATI. It's been shown over and over that the NV3x parts were slower in PS 2.0 than the equivalent parts from ATI.

-FUDie

 

[FUDie]

Your logic is flawed (what's new?). The R3x0 based cards support _PP. Do shaders with _PP modifiers suddenly cease to work on R3x0 cards? No. Since the R3x0 cards were faster running 24-bit precision than the NV3x was running _PP, your "argument" doesn't hold water (as usual).

Yeah, right. And all FX cards do support anisotropic level 16 then: if you set 16 in the API you'll still get anisotropic!

R3x0s DO NOT support _pp hints. They just IGNORE them. That is a big difference.

Sorry, but ignoring a request and giving higher quality is better than ignoring a request and giving lower quality. Hardly comparable. NVIDIA's caps settings for the NV3x series says a maximum anisotropy of 8, if the app tries to use more than that the app is in error.

_PP is a hint not a requirement. It's a hint from the app that the HW is allowed to use lower precision, it's not required that it do so.

Your arguments are about as valid as radar's... That's not a compliment.

-FUDie

 

[Tim Murray]

Note that radar has still not responded to Dave's statement that FP16 offers less precision than, for example, FX16.

Dave, out of curiousity, why hasn't "being an unrepentant moron" been added to the "Top Ten Reasons For Me To Ban Your Ass?"

(Oh, and Radar is spewing nonsense about ATI support for _pp = more SM2.0 games. the problem is that even if ATI DID support _pp and it offered a performance benefit, NVIDIA wold STILL have been too slow, even with _pp, to make widespread SM2.0 usage viable. Hell, I come back to this forum out of curiousity, and most of you have turned into inane fanboys who obviously have never even looked at the document that defines _pp.

And christ, people, R3x0/R420 does not support _pp. It's a hint that allows the hardware to use s10e5 as the minimum precision instead of s16e7. _pp hints can be ignored, and the hardware still complies perfectly to the DX9 spec.

oh, and what's the deal with PS1.4 on NV cards? doesn't NV's FX16-equivalent have a lower dynamic range than on the 8500, and in reality, it should be using FP32 at all times for PS1.4 suppport? Okay, I'm done editing.)

 

[andypski]

It shouldn't be difficult for an ATi supporter to modify a game to be pure DX9 though, just replace all 1.x shaders with 2.0 shaders.
I'm sure your hardware will cope, the ATi supporters are fond enough of telling us how good their DX9 support is.

Our hardware will certainly cope with no problems at all if you replace earlier shader models with PS2.0 shaders that do the same task. In fact, they'll even look pretty much exactly the same, because our hardware runs 1.x shaders at high precision all the time.

Some other vendor's hardware might not have such developer-friendly characteristics.

Of course, this begs the question, why do devs use PS1.4 in R3xx paths if the above is true?

Certainly not for any overriding performance reason. Logically it would seem to be so that they can write less shaders - if the effect can be conveniently expressed in PS1.4 then it can also be used for other paths and older cards (such as Radeon 8500), whereas if a 2.0 version is used then you need to have extra shader code and fallbacks for all earlier cards.

Pretty simple really. I'm surprised that you didn't think of it yourself. Oh, hang on a minute, maybe I'm not surprised.

Of course, some other cards that have problems with running full PS2.0 shaders at acceptable speed will be benefitting in performance terms from the use of PS1.4, since they will implicitly be able to compile many of the instructions to run at lower precisions.

If all DX9 graphics cards supported _PP, you would see more titles using SM2.0 and less using PS V1.4. Like I have said before, the lack of _PP on all cards has held back the adoption of DX9.

That's laughable in so many ways that I hardly know where to start.

We aren't forcing developers to code at full precision, our hardware doesn't suddenly stop working if the _pp flag is specified for instructions, and it doesn't run any slower. As a result developers are free to use _pp as much as they want (and as much as they can get away with considering the precision loss) knowing that it won't hurt our performance at all.

Your contention above makes no logical sense whatsoever.

 

[Bjorn]

If all DX9 graphics cards supported _PP, you would see more titles using SM2.0 and less using PS V1.4. Like I have said before, the lack of _PP on all cards has held back the adoption of DX9.

Now that is some funny shit.