TEV a keeper or a loser?

[MDX]

Now that more 3rd party developers are getting to understand how to make use of the TEV unit inside Nintendo's consoles:

Codemasters - Overlord: Dark Legend:

(Visual effects) There’re a couple of things that we did to help, there’re separate processes like the really good shadowing system on there. It would go, generally, with “next-gen” games and PC hardware, but we matched to get it working on the Wii hardware. We’ve tried to make it pretty much for the environment, to give that rich and luster feel. (Shadows from trees and leaves) Walker: yes, our real-time shadow system, that’s exciting, different; it gives us a really nice look. We’ve got bloom effects, 20 minions, 10 enemies on screen. We’ve worked to push the hardware, and there’s a lot going on.

(Use of the TEV unit, we saw advanced effects on GC, nobody using on the Wii) Yeah, we did use the TEV stages, which is essentially the graphic hardware of the Wii has. We did that quite a lot, especially to integrate: we had some real time code which generates them, so the shadows could work with that material like on the fly. That’s it, the way that we managed to get our interesting shadowing system was by manipulating the hardware with the TEV stages. (From scratch) It’s designed very much for the Wii hardware, it’s a custom code written specifically to get high performance out of the hardware.

High Voltage Software - Quantum 3 games

High Voltage Software claims that Quantum3 is the culmination of fifteen years of experience. It can create better visual improvements through a "16 TEV stage material pipeline that can blend up to 8 separate texture sources." In addition to supporting features that create better graphics and smooth gameplay, the technology also helps them do faster prototyping and content development.

Should Nintendo keep the TEV for their next console?
Why or why not?

 

[Shifty Geezer]

No, because it's outdated and inferior to contemporary programmable shaders. The only purpose is BC, which is unimportant. The fact deveopers can work around it's limits isn't a reason to keep them! If the hardware wasn't so esoteric, you'd get good looking results from all the developers instead of just a few. Heck, they shouldn't even have kept Hollywood for Wii, instead going with a conventional DX9-class GPU, let alone going forwards!

 

[sfried]

It should be built upon and improved. The interesting architecture leads to some very impressive results when used well. I'd much rather see them take this somewhere than just end up being homogenized into the old nVidia chipsets routine.

 

[Shifty Geezer]

Why? What's wrong with completely programmable shaders that let you do anything you want? The only benefit of weird hardware is when it can a performance advantage for being specialised which is a non-issue given the graphics algorithms we using these days.

 

[sfried]

Why? What's wrong with completely programmable shaders that let you do anything you want? The only benefit of weird hardware is when it can a performance advantage for being specialised which is a non-issue given the graphics algorithms we using these days.

That's why I said "build upon and improve". If you can have TEV shaders that are fully programable, who knows what kind of crazy effects you can use. I just don't see why it has to be DX9 compatible (which wouldn't even make sense considering Microsoft owns Direct X). If it were up to OpenGL 3.0 standards, I'm fine with that.

 

[Shifty Geezer]

The TEV is a fixed-function texture combiner unit. If it's fully programmable, it won't be the TEV, it'll be a (collection of) shader pipe(s). It doesn't have to be DX compatible* - my reference to DX was only the standard of GPUs out at the time Wii launched. Alternatively they could be described as SM2.0. Both descriptions give an overview of the hardware feature-set of the GPUs.

Intead of including a couple of fixed-function texture combiners, Nintendo could have included a handful of programmable texture combiners and ALUs (pixel shaders) that are far, far more versatile. Also using an off-the-shelf GPU would have allowed developers to use the ATi or nVidia development environments and languages, well understood by the majority of devs with any PC experience, and with a huge amount of reference material for anyone new to conventional GPU programming. The effort to learn and write powerful, versatile graphics code on an ATi/nVidia GPU in Wii would be a lot less than the effort to make the TEV do stuff it isn't well suited for, and it would produce better results. The only reasons not to have used such a GPU is for backwards compatibility and profitability.

TEV was a stop-gap solution, a cost/functionality compromise for last gen to get reasonable performance at reasonable cost. Sony's approach was massive bandwidth and fillrate and overdraw up the wazoo. MS's was very expensive versatile performance. As manufacturing technology has progressed and graphics algorithms and techniques have become more sophisticated, there's only one solution - high programmability. This is essential for the future. We're talking about unified shaders these days, or even CPU's powering graphics! Anything built along those lines will not be TEV, any more than RSX is an improvement on the Emotion Engine. There is no future in the GC GPU line. The only room for GPU innovation now is deferred rendering and eDRAM options. I can't see any options for realistic custom hardware, and even if Toshiba or Sony develop a GPU for PS4, I expect it follow the programmable shader/CPU design.

* If someone wrote a driver for it, it would be DX compatible, which would be pointless.

 

[Tahir2]

It was understandable for GameCube but for Wii and going forward it really is/was woefully outdated.

DirectX7 type graphics is not something to be built upon, but to be discarded! Like totally!

 

[MDX]

No, because it's outdated and inferior to contemporary programmable shaders. The only purpose is BC, which is unimportant. The fact deveopers can work around it's limits isn't a reason to keep them! If the hardware wasn't so esoteric, you'd get good looking results from all the developers instead of just a few. Heck, they shouldn't even have kept Hollywood for Wii, instead going with a conventional DX9-class GPU, let alone going forwards!

Fair enough, can you speculate if the transition would be easy from TEV to another solution for Nintendo First party and possibly second party developers? Considering they probably have been making use of it for the last two generations and may have all their resources and knowledge based on it.

And in terms of backwards compatibility, will it not be a problem for Wii games as they are more complex for the future console?

 

[tongue_of_colicab]

I suppose if you only ever programmed for TEV that it would be hard to jump straight into whatever next generation hardware has to offer. But I assume Nintendo isnt that stupid either and im sure they they will realize soon enough that if their teams dont have the knowledge needed that they get extra education and/or hire new personal that do have the experience needed.

DOubt backwards compatability is a problem given that there is already a working wii emulator that can run games fast so power probably isnt the problem unless nintendo decides to again build a machine only 2 to 3 times faster than wii but this time with a complete different architecture.

 

[assen]

The ease of transition is not interesting, since virtually nobody is doing advanced graphics on the Wii. Two or three internal Nintendo teams plus two or three third-party teams against the sea of DX9 SM3-level console developers.

As discussed many times before, the issue with advanced Wii graphics is not technical, it's economical.

 

[MDX]

Ok, what about some other factors like differentiation?

Does the TEV make Wii or Nintendo games look or feel different?

Or
is Nintendo trying to control the number games ported to their console?

Or
is Nintendo trying to get developers to get used to their way of working so developers become "committed" to their hardware?

Does keeping the TEV accomplish any of this?

But more importantly... why not both? Can Nintendo keep the TEV and include contemporary programmable shaders for their next console?

 

[Tahir2]

What is TEV?
http://forum.beyond3d.com/showthread.php?t=40460

An old thread with some nice posts by people who know what they are talking about.

The TEV and "pixel shaders" are basically cute acronyms for what used to be called color combiners. The TEV also incorporates the Texture reading part of the pipeline.

A color combiner is in general implemented as a single logic op, in NVidia's case thats public (register combiner docs) and is of the form
A op1 B op2 C op1 D
where op1 is either Dot Product or multiply, op2 is either add or select.
As you can see by repeating this multiple times with some register manipulation between stages you can do most basic math. Pixel shaders just provide a simple consistent interface to this (and other vendors implementations).

The TEV uses a different basic combine operation which is a little more limited. However since the Texture reads can be interleaved with the combiner operations it allows you to do things that would require multipass render on NV2X.

So as an example
on NV2X I have to write


Texture Read
Texture Read
.
.

Combiner Op
Combiner Op
Combiner Op
Combiner Op
.
.
.

On Flipper I can write

Texture Read
Combiner Op
Combiner Op
Texture Read
Combiner Op
Combiner Op
Texture Read
Combiner Op
Combiner Op
.
.
.

I guess the easiest explanation is that Flipper has simpler units for combining and reading textures, but allows more complex arrangements of the units.
So if one of the texture reads is dependant on a previous combiner Op and you can't squeeze the ops into the texture addressing instructions the NV2X would require multipass to do the same thing.

 

[Tahir2]

Flipper, the original GameCube graphics processing unit was designed for a console released in 2001. If Hollywood is virtually identical to Flipper then some of us are evangelising 8 year old technology!

More info on the TEV:
http://forum.pcvsconsole.com/viewthread.php?tid=9616

At that time the original XBOX GPU was NV2A which is based around the GeForce 4 architecture and has its Pixel Shaders and Vertex Shaders (versions 1.2 I think).

Vrey old stuff and all it means is that Nintendo wanted to spend little money on a new GPU for Wii but probably did want to:
a. hit a particular pricepoint
b. particular level of integration within Hollywood MCM (GPU+Northbridge+Audio+Memory)
c. certain power output (TDP) for form factor
d. ease of porting over tools to next console

We should be grateful we are getting some developers tap into the power of the GPU in Wii and the almost mythical TEV but honestly.. it gets the job done but it's pants compared to current tech. Since it was never meant to compete with current tech lets not overly get misty-eyed over it.

Probably the wrong thread but I am not that impressed with the development of the new games that may be responsible for all this TEV talk... GameCube was doing some amazing things many years ago as well (Factor 5, RE et al).

As to the question Ninendo trying to control ports to its console?

It may be an indirect effect and Nintendo really has only been concerned with 1st Party games (and lesser extent 2nd party) as software is really their cash-cow but then a console cannot survive on 1st party games alone. A bit of a catch-22.

 

[pjbliverpool]

But more importantly... why not both? Can Nintendo keep the TEV and include contemporary programmable shaders for their next console?

By my limited understanding, that would be a bit like trying to drive to work in a Ferrari and a unicycle at the same time.

In other words, they are two solutions to the same problem but one is far, far better in every respect than the other. To add a TEV alongside current (or even early DX9 level) GPU's would be utterly pointless.

 

[fearsomepirate]

Heck, they shouldn't even have kept Hollywood for Wii, instead going with a conventional DX9-class GPU, let alone going forwards!

Yeah, if they'd gone with an off-the-shelf DX9 GPU, maybe the Wii would have been successful.

The only reasons not to have used such a GPU is for backwards compatibility and profitability.

What a silly reason. Nintendo's execs must be green with envy when looking at what the PS3 has done for Sony.

Anyway, it really depends on what they want to do with their next console. I am expecting something DSi-like, i.e. something that is basically an upgraded Wii with more performance and features, featuring "enhanced" software that will work on both machines. If that's the case, then the GPU needs to be 100% with the current one, meaning some version of the TEV will be in it.

If they're going for a clean break, then there's no need.

 

[Shifty Geezer]

What a silly reason. Nintendo's execs must be green with envy when looking at what the PS3 has done for Sony.

No need for the sarcasm! I'm not saying it's a bad business call, but in a tech thread asking if progressing TEV is a good idea, the answer is a categorical 'no' and Nintendo's choices were not in any way based on the technical merits of the hardware.

 

[pjbliverpool]

Saying that though, even a low end DX9 GPU (X1300 level-ish) would have been more than capable of playing the current Wii games at HD resolutions.

Its possible with the selling point of being HD that the Wii could have done even better. I have no idea how Hollywood compares to such a GPU in price terms though.

 

[fearsomepirate]

but in a tech thread asking if progressing TEV is a good idea

What makes a technological decision a "good idea" or a "bad idea" depends on the goals of the party making the decision. You can't just ask the question in a vacuum. For example, is an ARM CPU "good" or "bad"? Well, it's "good" if you're making a mobile device, and "bad" if you want to run the latest Crytek engine.

Obviously, a fixed-function GPU has no advantages over modern shader hardware if the question is solely one of features and capabilities. But if that's the only question, the answer is that Nintendo should start making PCs with removable graphics cards.

the answer is a categorical 'no'

No, it's not. You have to qualify your 'no' to only cover the question of hardware capability. Since there are exist contexts in which continuing the current technology would be appropriate, 'no' can't be said categorically.

Nintendo's choices were not in any way based on the technical merits of the hardware.

Reusability and power consumption are technical merits. Iwata was very explicit that one design goal of Wii was to minimize power consumption and physical profile (a low-end DX9 GPU still has far, far more transistors than Wii's GPU), and early SDKs emphasized the ability to simply re-use Gamecube code. However, given that most people on this forum discuss Nintendo's business strategy while overtly ignoring what Nintendo executives have said about the subject, I suppose that what their executives have said about their hardware decisions is irrelevant as well.

Its possible with the selling point of being HD that the Wii could have done even better.

Negligibly. People who care about hardware specs would still have mostly bought the far more powerful machines, and people who mainly care about violent games still wouldn't have been interested. Nintendo would have only grabbed a portion of the 'tail' of graphics-oriented customers, who aren't even that numerous to begin with.

 

[Shifty Geezer]

No, it's not. You have to qualify your 'no' to only cover the question of hardware capability. Since there are exist contexts in which continuing the current technology would be appropriate, 'no' can't be said categorically.

You'll have to convince me that there's a sane category in which fixed-function TEVs make sense in a Wii2 piece of hardware.

Reusability and power consumption are technical merits. Iwata was very explicit that one design goal of Wii was to minimize power consumption and physical profile (a low-end DX9 GPU still has far, far more transistors than Wii's GPU)

So create a different, smaller GPU that is still programmable! Also why is power consumption such an issue in Wii? Whether it's 20 watts or 40 watts makes no odds. Giving drawing a few more watts to get substantially better graphics is a choice I think pretty much everyone would make.

...and early SDKs emphasized the ability to simply re-use Gamecube code.

Okay, but you can't be backwards compatible forwever. Technology moves forwards too fast. If Nintendo really want to plug in existing code, they can design a flexible GPU that supports Broadway instructions but adds full programmability so developers who want to use new techniques instead of decade old ones can.

However, given that most people on this forum discuss Nintendo's business strategy while overtly ignoring what Nintendo executives have said about the subject, I suppose that what their executives have said about their hardware decisions is irrelevant as well.

You're just being shirty here. No need to make sweeping statements about reader attitudes. Present your argument, not an editorial on reader attitudes

Negligibly.

Debateable. It's not just about paper specs, but what you see on screen. When someone sees a Wii on a large HDTV and says 'gosh, that looks a bit rough' then you know Nintendo aren't hitting every box and there's room for improvement that should lead to more sales. We can never no how many people would have been more attracted to Wii had it a better GPU, but it's certainly some people. Also a standard GPU would have meant better visuals on existing games which could well have led to improved sales.

But that's going off topic. On topic, what are the reasons for keeping fixed-function TEVs next-gen? I say there's a categorical 'no'. Prove me wrong with a convincing example of a games-system that'll be competitive while being tied to decade old technology!

 

[Tahir2]

On topic, what are the reasons for keeping fixed-function TEVs next-gen? I say there's a categorical 'no'. Prove me wrong with a convincing example of a games-system that'll be competitive while being tied to decade old technology!

Only one I can think of is backward compatibility of development software (and also user games) - which could save Nintendo millions.

How about a Hollywood that was redesigned (still fixed function) but took a quad GPU core approach. So we have 4x, or even 8x, Hollywood with complete compatibility with older games. Any good for the consumer and developer? I would still say a categorical no.

Why is Hollywood lacking?

Programmability limitation - 16 register combiners limits flexibility
Frame buffer limitation
Lack of FP support limiting accuracy (e.g. HDR)
Lack of performance for AA
Texture size limitation
Fixed function TnL engine limitations
32bit wide access to main system RAM (64MB GDDR3)

The actual core logic of Hollywood is probably closer to 20 million transistors (excludes the 3MB of texture and z buffer RAM).

If Nintendo kept the TEV, added FP10 and FP16 formats to it, added another 12 pixel pipelines capable of multitexturing and increased the register combiners from 4 per TEV pipeline to 16 the architecture (256 altogether) and then increased the core clock to increase pixel fillrate, made the texture ram redundant/increased the texture cache and speed, added MSAA support in hardware, Hollywood would begin to resemble DirectX 9 equiavalent hardware anyway, wouldn't it?

Just put an R300 in there...