Criticism of Nvidia's TWIMTBP Program - HardOCP's Just Cause 2 Review

[newtekie1]

You have misunderstood me. I did not state that AA is something that exists in all graphics engines, but rather something that should exist in all graphics engines. Where we differ is that I place onus on the developer to ensure that this feature is included, irrespective of the engine they decide to use.

Yes, this is my thing as well. If they cannot do it, they shouldn't be developing games in the first palce. Basic stuff, AA. Next, they won't be able to get AF to work...

I mean really...you can expect very little from those developers, for sure. If it was any other character than Batman, the game would have been passed over by everyone that hyped it as being so good.

Game developers use pre-made engines for a reason. It is cheaper and less time consuming then creating their own. That certainly does not make them not suited to develope games. If we are saying that every game developer that uses a pre-made game engine shouldn't be making games, we wouldn't see 90% of the games we have, including some extremely good ones.

It takes huge resources to develope an engine, and it takes huge resources to implement such things as in game AA. Which is exactly why most game developers use another larger studio's pre-built engine. We are talking about some pretty big developers too: EA, Midway, Sony, Square Enix. Some pretty good games also: X-men Origins, Unreal Tournament 3, Rainbox Six Vegas and Vegas 2, Stranglehold, Mirror's Edge, and the list goes on.

Hell, even Unreal Tournament 3, the premier game on the Unreal Engine 3 lacks in game AA.

Uh, you owe me an apology.



http://ati.amd.com/products/radeonx1900/index.html



http://ati.amd.com/products/radeonx1800/index.html

Every card since the X800 has supported PS3.0. Lack of support on legacy products in unimportant.

No, just because you have an inability to read and comprehend, that does not mean I owe you and appology.

dave he has ony 3 ### meaning 850/800 or similar cards. At this time the 6800/7series already had it.

none of my X850 XTPE's has SM3....didnt run the putry dragon test in 3DM06 till I got a 7600

He said X### NOT X####. That is X800, X600... Nvidia had PS3.0 on their hardware GF6800, ATi was like 6 months late with PS3.0 hardware. Games like Oblivion lacked proper PS3.0 for that reason, for example.

See, they don't have a problem with reading and comprehension.

 

[HeroPrinny]

chill people chill, this could be a very interesting topic but this. back and forth slagging ruins it.

 

[cadaveca]

Hell, even Unreal Tournament 3, the premier game on the Unreal Engine 3 lacks in game AA.

Actually, only in DX9 does it not support AA, due to using deferred rendering. It's the use of deffered rendering that makes it THE choice for games with Phys-X. There's no reason for AA to be left out of DX10 UE3 games.

Tim Sweeney: Yes, we'll ship Unreal Tournament 3 with full DirectX 10 support. Support for multisampling is the most visible benefit.

 

[Benetanegia]

Uh, actually, it was. Back then it was called WGF, and was part of "Longhorn". Circa 2001. DX 9.0c released in 2004, 3 years after work on DX10 began.

http://en.wikipedia.org/wiki/Development_of_Windows_Vista

http://en.wikipedia.org/wiki/DirectX

Vista is not that Longhorn, in fact Vista was more of a rushed Win7 than anything else. The original Longhorn was delayed and much changed over the time, eventually adopting many things from the next OS M$ was working on, now called Win7. Things are not black and white, so M$ released 9.0c, because Longhorn was no longer what it was thought to be: a OS for 2003 release. They included things that were supossed to be part of Longhorn? Obviously they could not wait until Vista (aka rushed Win7) was released, so they implemented things that were much needed. They crippled DX10 because that? Nope and certainly Nvidia had nothing to do with that, which is what my post was about anyway.

 

[EastCoasthandle]

You know what's interesting about that review is that even with those IQ features disabled the 480 still loses. I read some of the posts over there and one person said that the benchmark test clearly put the 480 ahead. I wonder if the drivers were optimized for the benchmark?

 

[cadaveca]

Vista is not that Longhorn, in fact Vista was more of a rushed Win7 than anything else. The original Longhorn was delayed and much changed over the time, eventually adopting many things from the next OS M$ was working on, now called Win7. Things are not black and white, so M$ released 9.0c, because Longhorn was no longer what it was thought to be: a OS for 2003 release. They included things that were supossed to be part of Longhorn? Obviously they could not wait until Vista (aka rushed Win7) was released, so they implemented things that were much needed. They crippled DX10 because that? Nope and certainly Nvidia had nothing to do with that, which is what my post was about anyway.

We can start another thread on this all on it's own, there's so much to this subject. So I'll leave the O/T chat for now. Your impressions of what longhorn, WGF, and DX10 are is a bit skewed, but if you can put up with me, I'd more than love to talk about it in another thread.

 

[bpgt64]

The developer is only responsible to only deliver a product upon which the vast majority of their target audience will be willing to purchase, and thus sustain that company until next release. If you take issue with the product, vote with your wallet. Just like the Graphic's Card company has to measure feature(s) vs cost. In the end, the goal is simply to make profit.

We, are the minority that spends heavily(in comparison to the average), we demand more and drive the industry. If Nvidia wants to insure there product has a greater feature set, and yes paying a developer to test code is a valid way to do so. Then is not Nvidia's fault by any-means to do so. Where things get dicey, is when a company can pay to put a competitor at a disadvantage intentionally. That, if provable(which conclusively has not been or ever will be proven in a forum such as this), goes against Anti-Trust(mainly anti-competition) laws. Like the ones Apple is about to face, despite being a minority provider.

Now, your thinking well paying a developer to test there code on your hardware by definition puts an apposing GPU manufacturer at a disadvantage. No it doesn't, Ati had a choice, and it made its decision. Again, vote with your wallet if your unhappy with your horse.

 

[Benetanegia]

There's no reason for AA to be left out of DX10 UE3 games.

I think we can all agree with that, although I'm not sure they did implement it in the end. And there's no reason to leave it out of DX9 either. It can be done after all and Batman is the proof. Fact of the matter is that all and every UE3 game released to date is a DX9 game, as so is 99% of the games released (copy/paste the code and create a DX10 executable doesn't make it a DX10 game). It's the choice that developers are making and only them are to blame. Nvidia's intervention (or AMD's in the case of the few games where they helped) is always a good thing for us, consumers, gamers.

 

[cadaveca]

Sure, but my point is that in DX10 UE3, AA is natively supported(on DX10 hardware).

Q: So why did nV need to code it?

A: For DX9 cards.

Q: So why does that effect ATi cards in DX10?

A: Because the developer purposely changed the AA implementation throughout the engine, breaking AA on ATi's cards, using nV's code.

So, nV broke AA in Batman, and seemingly, purposefully. This is the real basis for the whole argument in the first place.

 

[Benetanegia]

Q: So why does that effect ATi cards in DX10?

A: Because the developer purposely changed the AA implementation throughout the engine, breaking AA on ATi's cards, using nV's code.

Once again no. DX10 supports AA with deferred shading, but UE3 does not. Plain and simple. They said they would implement it on DX10, but I highly doubt they did in the end. And all that is irrelevant, since every game is using the DX9 code anyway (is there even a DX10 code path for UE3?). My point is it doesn't matter if Epic implemented AA on DX10 UE3 or not (if it exists at all), because developers are buying and using the DX9 one, for multi-platform purposes.

 

[cadaveca]

Yes, there is a DX10 codepath, and I know for a fact that UE3 supports AA.

Tim Sweeny who I quoted above, is the engine's writer, by the way. I know for a fact that since early 2008 AA has worked in DX10...you can even enable it in Vista with the UT3 demo.

And they do not "just buy the DX9 version"...one of the major selling points of UE3 is that you need not worry as the engine will scale features automatically to the platform it's on. Nv's code changes this detection method to use nV's code for AA at all times, rather than defaulting to the normal path for AA.

I mean sure, nV's code is brilliant. It does come at very little performance penalty compared to running the native implementation, but that doesn't chagne that fact that the code customized the engine, and broke AA.

I can even explain why the native method comes at such a performance hit, but that's irrelevant.

 

[Benetanegia]

Yes, there is a DX10 codepath, and I know for a fact that UE3 supports AA.

Tim Sweeny who I quoted above, is the engine's writer, by the way. I know for a fact that since early 2008 AA has worked in DX10...you can even enable it in Vista with the UT3 demo.

And they do not "just buy the DX9 version"...one of the major selling points of UE3 is that you need not worry as the engine will scale features automatically to the platform it's on. Nv's code changes this detection method to use nV's code for AA at all times, rather than defaulting to the normal path for AA.

I mean sure, nV's code is brilliant. It does come at very little performance penalty compared to running the native implementation, but that doesn't chagne that fact that the code customized the engine, and broke AA.

I can even explain why the native method comes at such a performance hit, but that's irrelevant.

I would like you to show proofs of that, because I was never able to use AA on UT3 (nor Mass Effect, nor Mirror's Edge, Borderlands, Brothers in Arms....), except via Control Panel, something that you can do with Batman too. I'd really apreciate it, because I see many people claim all those things everywhere and I have yet to see a single proof of anything.

 

[FordGT90Concept]

http://ati.amd.com/products/radeonx1800/index.html

Every card since the X800 has supported PS3.0. Lack of support on legacy products in unimportant.

I started the X### thing here.

9### = 9700 and 9800 family of cards (DirectX 9.0, Pixel Shader 2.0)
X### = X800 and X850 family of cards (DirectX 9.0b, Pixel Shader 2.0b)
X1### = X1800, X1900, and X1950 family of cards (DirectX 9.0c, Pixel Shader 3.0)
HD 2### = HD 2900 family of cards (DirectX 10.0, Pixel Shader 4.0)
HD 3### = HD 3800 family of cards (DirectX 10.1, Pixel Shader 4.1)
HD 4### = HD 4800 family of cards (DirectX 10.1, Pixel Shader 4.1)
HD 5### = HD 5800 family of cards (DirectX 11.0, Pixel Shader 5.0)

The X### Pixel Shader 2.0b cards debuted after GeForce 6 series which supported Pixel Shader 3.0. X1### didn't debut until late 2005, after the introduction of GeForce 7 series. ATI was an entire generation behind NVIDIA at that time.

 

[cadaveca]

I would like you to show proofs of that, because I was never able to use AA on UT3 (nor Mass Effect, nor Mirror's Edge, Borderlands, Brothers in Arms....), except via Control Panel, something that you can do with Batman too. I'd really apreciate it, because I see many people claim all those things everywhere and I have yet to see a single proof of anything.

Well, the DX10 AA uses SuperSampling AA. As we all know, not all drivers support this properly, so that may be a major source of people not getting working right. It's also why it comes at such a huge performance penalty.


This is why you can't get it working in those games...they've simply turned off the switches in the GUI to enable it, but beleive me, it's natively supported by the engine. Current gen cards definately have the power to use SSAA, but mid-range and lower cards might not be so capable.

What it boils down to is that deffered rendering is hard to balance properly for most programmers. So they have to make choices as to where they spend their dollars, as many have mentioned, and AA is low on this list, due to the aforementioned performance problems.

However, deffered rendering is the way of the future for all engines, so programmers better get used to it, and get good at it too...and when things happen like nV stepping in, this prevents the need for them to do so. THAT's why I have issue with nV and them helping with AA...it's cool that nV users get a benefit, but it should not break the experience for others.

 

[Benetanegia]

Well, the DX10 AA uses SuperSampling AA. As we all know, not all drivers support this properly, so that may be a major source of people not getting working right. It's also why it comes at such a huge performance penalty.


This is why you can't get it working in those games...they've simply turned off the switches in the GUI to enable it, but beleive me, it's natively supported by the engine. Current gen cards definately have the power to use SSAA, but mid-range and lower cards might not be so capable.

What it boils down to is that deffered rendering is hard to balance properly for most programmers. So they have to make choices as to where they spend there dollars, as many have mentioned, and AA is low on this list, due to the aforementioned performacne problems.

Yeah SSAA has always been an option for UE3 (always through CP) and it also is available in BM:AA. It's MSAA what it lacks and what all this thing is about. While not exactly MSAA, Nvidia's implementation is similar and that's why it doesn't have as large a penalty as SSAA. And that was always the thing, that you had to enable SSAA through CP for Ati cards, but Nvidia had this MSAA+edge detect thing. SSAA has always been available and yes it's something anything DX10 can do because it's not really part of the fragment stage of the renderer, but that's not what has ever been discussed. I think you are a little bit lost here. So much written and you didn't even understood the beginning of the story. SSAA is there for all cards, "MSAA" is not.

 

[cadaveca]

Yes, but the engine has the ability to enable SSAA with the console...there's no need to use the CP to enable it, other than developer's choice.(works with the UDK).

I get what you are saying, but that's not the point. SSAA is broken too, in Batman. It should not be.

 

[ShiBDiB]

mods shoulda listening to me in the beginning... it didnt turn into a nvidia bashing thread, but an E-Peen war on who knows more about AA

 

[sneekypeet]

so? its still civil, quit poking or you will earn yourself points.

Either post on the topic or move along, thank you.

 

[BababooeyHTJ]

Well, the DX10 AA uses SuperSampling AA. As we all know, not all drivers support this properly, so that may be a major source of people not getting working right. It's also why it comes at such a huge performance penalty.

The quote that you posted by Tim Sweeny mentions Multi-sampling, btw.

 

[cadaveca]

The quote that you posted by Tim Sweeny mentions Multi-sampling, btw.

Yes. And what is the difference between MSAA and SSAA?

 

[Benetanegia]

Yes, but the engine has the ability to enable SSAA with the console...there's no need to use the CP to enable it, other than developer's choice.(works with the UDK).

I get what you are saying, but that's not the point. SSAA is broken too, in Batman. It should not be.

It's the first time I hear that SSAA is broken for BMAA. It certainly isn't a widespread issue (0 google resluts on this end).

Maybe:

As we all know, not all drivers support this properly, so that may be a major source of people not getting working right. It's also why it comes at such a huge performance penalty.

Many games have issues with SSAA, so I'm not doubting that BM:AA has some. I've had some issues with many games myself.

In any case the point comes home that Nvidia has nothing to do with the lack of proper AA in BMAA (in fact, if SSAA is broken the implementation of their AA is a blessing). And neither does with any of the many other things they've been accused for. But I guess it's easier to blame someone, even if it's not the correct one...

Yes. And what is the difference between MSAA and SSAA?

Are you serious?

 

[ctrain]

Your assumption that AA is something that exists in all graphics engines and game titles by default is wrong.

of course it exists by default in all graphics engines, the question / problem at hand is rather if any of the rendering techniques they use breaks hardware aa...

...like the case is with UE3, it's a DX9 limitation

 

[cadaveca]

Are you serious?

No, but if you understand the differences, you know that Tim was refering to SSAA only. DX10.1 is required for MSAA in deferred rendering.

Well, at least it was, and that's what makes nV's code so great.

Why that screws with SSAA, I don't know the precise details of, but i do have a couple of good possibilities. I need an nV card to find out the exact cause.

If it is using DX10.1-type commands...wow...there really is something to this...because according to nV, thier card's don't support 10.1. So how the heck did they get that working?

 

[enaher]

Well the use of proprietary technology sucks big time, but Nvidia's TWIMTP, is not to blame, Nvidia offers extra's and that's fine, I can't see nothing wrong with it, actually Id like Ati to work with Devs, and improve Stream and Opencl applications and promote open standards, I'm sure Ati hardware has some untapped potential.

 

[Benetanegia]

DX10.1 is required for MSAA in deferred rendering.

It was always posible via a separate Z buffer (even on DX9), although was more tricky and maybe not as desirable. DX10.1 only made it faster and introduced a clearer interface.

edit: Yeah it wouldn't be exactly MSAA, but rather an edge detect algorithm + SSAA on the edges. So... it wouldn't be a Coke, it would be Pepsi, but essentially the same. It's probably exactly what it's being done in BM:AA.