TEV a keeper or a loser?

[Nano]

To be honest MDX, they would be able to keep it cheaper and many times more powerful by going for something more up to date. Even the cheapest integrated graphics chips nowadays dwarf the current capabilities and can output HD.

I think it would be too much trouble. I see the TEV as an early pixel shader, so for it to be a better piece of kit, it needs to be far more programmable and flexible, which means practically ditching it.

 

[fearsomepirate]

All else follows for power consumption because the chip would be the same size

No, it wouldn't have. The Gamecube's GPU had ~24m logic transistors. The smallest ATI DX9 chip is the Radeon X300, at 75m transistors--and with 4 pixel pipelines at 325 MHz, you're not doing anything too exciting in HD. And even then, at three times the area and 33% more clock cycles per second, it's going to be consuming significantly more energy than the Wii's tiny chip. If you're serious about HD and want those DX9 features to actually get used, you're going to need more pixel pipelines. You're looking at 100m transistors at the minimum and maybe 400 MHz, just by looking at the transistor counts of video cards in the 9700-X800 range.

New chip features are made possible because new fabrication processes allow engineers to put more transistors on a chip and clock them at higher speeds, not because they simply rearrange the same number of transistors over and over in ever more clever ways and find ways to make four pixel pipelines draw extra pixels on a single clock cycle. Where are people getting this idea that you can design (with apparently negligible man-hours!) a chip with all the features of a current-gen GPU with the transistors of a last-gen GPU?

lso would they really need flipper that much for BC? There is a fairly decent wii emu for pc already.

Software does not write or debug itself. The question is whether or not it's worth paying people to write an emulator, and whether that additional cost will make more customers, not simply whether or not it can be done.

Also, retaining the old architecture means minimal cost in updating SDKs.

Note to all:

I am not saying Nintendo can or should keep the same architecture. I'm simply saying that it depends on their business plan, and I can imagine a business plan for following up Wii where that would make sense. I've noticed not one of you seems to have anything to say about what Nintendo's business plan ought to be in order to follow up Wii. If you don't have an "ought" for the business end, you really can't come up with an "ought" for the technology end, since Nintendo is not in the game of simply creating new, powerful silicon and finding ways to sell it, like ATI or Intel. Nintendo's been very clear about abandoning the technology-driven roadmap. Their roadmap is based on other considerations entirely, and they now choose technology to fit those plans.

 

[AlNom]

No, it wouldn't have. The Gamecube's GPU had ~24m logic transistors. The smallest ATI DX9 chip is the Radeon X300, at 75m transistors--and with 4 pixel pipelines at 325 MHz, you're not doing anything too exciting in HD.

For reference: the actual Wii GPU (incl eDRAM) is ~72mm^2 @90nm. The rest of the package is MCM including I/O or 1T-SRAM etc.

X300 is 84mm^2 @ 0.11um. But the point is feature capability. Nevermind HD, I don't care about HD. Yes, that is sans eDRAM, but the key is SD res. Ditch the eDRAM, and make for a wider path for main mem bandwdidth). Nintendo never intended for HD. Anyways, the point was programmable shaders which are more flexible. At least, I thought the thread was about features, not necessarily HD. Afterall... "TEV" was the point of contention. It should be very obvious that one can achieve more advanced features within the same die space. For the die size, a 24-bit framebuffer is just ancient. :s

But at any rate, seems a bit of a strange discussion. TEV from start of the century showing up more than a decade later Really...

 

[MDX]

Note to all:

I am not saying Nintendo can or should keep the same architecture. I'm simply saying that it depends on their business plan, and I can imagine a business plan for following up Wii where that would make sense. I've noticed not one of you seems to have anything to say about what Nintendo's business plan ought to be in order to follow up Wii. If you don't have an "ought" for the business end, you really can't come up with an "ought" for the technology end, since Nintendo is not in the game of simply creating new, powerful silicon and finding ways to sell it, like ATI or Intel. Nintendo's been very clear about abandoning the technology-driven roadmap. Their roadmap is based on other considerations entirely, and they now choose technology to fit those plans.

Nintendo has stated recently that their next console will be able to play games at HD resolutions.

I do really think Nintendo will take backwards compatibility very seriously not only for games, but all the peripherals the Wii uses. I've got a feeling that they will have to consider how best to keep their new fans from leaving them for the competition who, as we now know, will also be incorporating motion controls and will probably aggressively go after the casual market.

Based on everyone comments here, If Nintendo plans on dropping the TEV, I think the best solution for them is to include the Wii chip into the new hardware like what Sony did with the first gen PS3 for backwards compatibility. I mean how expensive would that be for them? Or I should say, would it be a problem to sell decent backwards compatible Wii-HD at a price along 250 in two to three years?

 

[swaaye]

Considering how sparse the library is with respect to quality games, they could drop backwards compatibility and I'd be fine with it. I don't need Cube compatibility either.

 

[MDX]

Considering how sparse the library is with respect to quality games, they could drop backwards compatibility and I'd be fine with it. I don't need Cube compatibility either.

*ouch*
I'm holding back from starting a list war.

Well all I can say that considering that many cool, decent, and well received titles are not selling that well on the Wii, backwards compatibility might give those titles a new life. They will be cheaper, and can offer some filler for those early droughts of a new console.

Consider the DSi. How many games, at the moment, are really designed only for it?

Anyway, thanks everyone for their opinion on this matter.
It will be interesting if this thread gets brought back to life once we know the innards of the new Wii 2.

 

[swaaye]

Well I think my favorite "Wii" game is RE4 Wii edition. Most of the other highly acclaimed games aren't my thing. I have given Zelda, Mario and Metroid a try. Hell, I put ~50 hours into Zelda but it never really did much for me and I just stopped playing. I think the new Madworld game is kinda dumb. I do enjoy multiplayer light gun games like RE UC, but they aren't exactly serious attempts at quality game programming. And there are zero quality racing games, not even a new FZero or ExtremeG like game (MarioKart = blah for me [best on SNES lolz]).

I still have a few Cube games but I've just losing interest in them and haven't played them more than a hour or so since I got the Wii.

The best thing about Wii for me is that I soft-modded it and it can now play emulators without having to buy the games again from the Virtual Console nonsense.

 

[fearsomepirate]

Nevermind HD, I don't care about HD.

Iwata's said the Wii successor will almost certainly have HD. The thread is about whether Nintendo should basically go with a version of the current fixed-function GPU for their next console, or use something more complicated. Going HD is a given. I think backward compatibility with Wii is almost guaranteed--you don't want 15m people feeling like they have to throw out their Wii Fit boards. They may even want to keep selling it, or some version of it. If you want to do those two things, you can revise the current chip, go with a two-GPU solution, program an emulator, or maybe go with a truly new chip with a built-in ability to interpret Flipper instructions.

But at any rate, seems a bit of a strange discussion. TEV from start of the century showing up more than a decade later Really...

The DS has 15-year-old graphical features. How big you can make the chip is ultimately depends on what the machine is for, which depends also on to whom you're selling it.

 

[Shifty Geezer]

Iwata's said the Wii successor will almost certainly have HD. The thread is about whether Nintendo should basically go with a version of the current fixed-function GPU for their next console...

HAng on, your line of debate is meandering a bit here. The point you made which AlStrong replied to was with regards the Wii GPU's HD performance...

No, it wouldn't have. The Gamecube's GPU had ~24m logic transistors. The smallest ATI DX9 chip is the Radeon X300, at 75m transistors--and with 4 pixel pipelines at 325 MHz, you're not doing anything too exciting in HD.

Wii doesn't need to do anything exciting in HD as it's an SD console. If you're going that route, you need tol compare X300 at 720p to Hollywood at 720p

I think backward compatibility with Wii is almost guaranteed--you don't want 15m people feeling like they have to throw out their Wii Fit boards.

Two counterarguments here. 1 - Wiifit is a peripheral, so no-one needs throw it out. They just buy a new piece of software that uses it. Nintendo can bundle the software if they wish. 2 - Do you really think Nintendo would rather every WiiFit owner uses their existing board instead of buy a new profitable peripheral for their new console?!

You'll need to present some convincing numbers if you want me to think BC is important to a substantial number of people. I personally don't think the Wii install base is made up of lots of GC fans who would not have migrated if Wii didn't play their existing GC library, and who won't buy Nintendo's next machine if it doesn't play their 10 year old GC library.

 

[fearsomepirate]

- Do you really think Nintendo would rather every WiiFit owner uses their existing board instead of buy a new profitable peripheral for their new console?!

Planned obsolescence doesn't make customers happy (and it's even more profitable to keep selling balance boards as long as you can). I kind of hold Wii fit up as the cardinal example of the new-type customer. What would a 29 year old woman who bought Wii Fit want a new console to do? Would she want to keep using Wii Fit, or would she be tired of it?

You'll need to present some convincing numbers if you want me to think BC is important to a substantial number of people.

Nintendo's not going the route of basing their future plans on the technological trends of the last two console generations, so I don't think that's actually very useful. In past generations, people primarily bought a new console for access to better graphics. Well, that's obviously not what's driving Wii sales, so you need to think about the values of its customer base. Looking at past data and extrapolating out trend lines is what brought us the PS3 and Xbox 360. While they are fine machines, that's just not how Nintendo is building itself.

 

[pjbliverpool]

I have to agree that its a different type of customer using the Wii to that which has traditionally purchased consoles. We are talking families, older people, people who for the most part, don;t care much about technology or graphics and simply want something they can have some family fun with.

Its kindalike the modern board game only a lot more expensive. I imagine that if a couple of years from now Nintendo were to tell all those people they need to buy Wii 2 with which all their previous games and peripherals will not work they would have a lot of dissillusioned customers.

Having said that, the Wii "Control Plus" or whatever its called has a similar effect and N jumped right on that. Personally, I think telling tens of millions of customers that those 2 - 4 Wii motes they spent £120 on just a few months ago are "no longer good enough" is a pretty dumb idea. Even as a tech head it makes me wonder if there is any point in committing to and investing money (peripherals, games) in a platform if the vendor is going to be so fickle about it.

 

[swaaye]

What happens is companies introduce "exciting new features" that the previous product didn't have and then new product is no longer a simple replacement for old product. It's the hot new "sequel remake edition". Suddently people become much more willing to "upgrade". Not exactly a new concept in consumer electronics and something that almost everyone accepts. It's like the basis for consumerism I think.... Like how I paid $250 for a Cube with more clockspeed, ultra cheap minimal RAM allotment, and a somewhat gimmicky control scheme!!!

I think the Wii market is jam packed full of noobs that are ready to be milked for all they're worth.

 

[Mobius1aic]

I think the Wii market is jam packed full of noobs that are ready to be milked for all they're worth.

Hardcore Nintendo fanboys and families

 

[forumaccount]

Better DOT3 support should be on the agenda if they do extend the TEV system. Dot3 in every unit? NV_register_combiners had that on GF2, yeah? But only 2 units + a final combiner.

Wii was quite a gamble for Nintendo. I think they got caught with their pants down by MS and Sony's release dates and didn't have time to do anything more complex than the Wii hardware we have now. Adding motion controls was a risk as well since they had experimented with this a number of times over the years and it was never well received. By some miracle, the package worked and took off... but I think it was more in spite of the limited hardware than because of it.

I don't know if they'd play the same trick again. One hopes they have made enough money on Wii and are planning far enough ahead to "do it right" next time. TEV is really pretty useless to bring forward except for backward compatibility, but emulation solutions are probably the way to go there. I don't know where people got the idea that TEV is some magical mystery box that holds tons of potential to create graphics that rival or surpass new consoles, but it's wrong. Anything TEV can do, Xenos or RSX can do (better?).

 

[DRS]

Better DOT3 support should be on the agenda if they do extend the TEV system. Dot3 in every unit? NV_register_combiners had that on GF2, yeah? But only 2 units + a final combiner.

It already has a dual dot3 in the texture unit. What's missing is the ability to transform a light to tangent space, normalize its vector and, per pixel, pass it to the dot product matrix. In contrast, if you have a light direction, you can set up the texture unit with it to perform normal mapping on directional light. But the CPU is required to transform it.

Besides that, for normal mapping you don't need dot3 at all, I'm currently working (it already works) on a fast, all GPU, normal mapping method. It just requires an indirect- and a tev stage to do so (though it results in some less appealing artifacts, which can be dealt with by adding an additional tev stage). I'll post my dol later on for non believers (probably on gcdev).

 

[Entropy]

It already has a dual dot3 in the texture unit. What's missing is the ability to transform a light to tangent space, normalize its vector and, per pixel, pass it to the dot product matrix. In contrast, if you have a light direction, you can set up the texture unit with it to perform normal mapping on directional light. But the CPU is required to transform it.

Besides that, for normal mapping you don't need dot3 at all, I'm currently working (it already works) on a fast, all GPU, normal mapping method. It just requires an indirect- and a tev stage to do so (though it results in some less appealing artifacts, which can be dealt with by adding an additional tev stage). I'll post my dol later on for non believers (probably on gcdev).

DRS, since you program the thing, and seem familiar with its innards, could you please make a completely personal comment on what you would like to see changed/updated on the GPU, assuming Nintendo choose that path for its next offering?
If you explain why and put it in some order of importance, it would be even better. The enlightenment would be appreciated.

 

[stiftl]

Wouldn't it be possible to just simulate the hardwired TEV with a fully programmable pixel shader? At least they know what's inside the box, but would it be possible from a technical view?

 

[Entropy]

Wouldn't it be possible to just simulate the hardwired TEV with a fully programmable pixel shader? At least they know what's inside the box, but would it be possible from a technical view?

Speaking as an amateur, it should be possible.
However, your question is purely theoretical. In practise, the questions that need to be answered are:
1.1 "Can they do it with guaranteed high performance while keeping within their desired envelope for cost, power draw and performance for new applications"?
1.2 "If so, is it really the best way to achieve backwards compatibility taking all other relevant factors* into consideration?"


* Relevant factors being, for instance cost and power, obviously, but also effort in tool creation, time to market, desired performance for new applications, et cetera.

PS. The question I would like to see discussed here is actually this one [note:backwards compatibility optional]:
2 "What are the options Nintendo can choose from, given reasonable constraints in cost and power, and what are the implied consequences of those choices?"

 

[Shifty Geezer]

Wouldn't it be possible to just simulate the hardwired TEV with a fully programmable pixel shader? At least they know what's inside the box, but would it be possible from a technical view?

Yes, every TEV stage should be fully translateable into a shader program as I understand it. I don't know if TEV has some funky memory addressing or somethng to spoil that up though.

 

[stiftl]

So, as I see it there wouldn't be any reason to keep the TEV and not move on with a pixel shader in the next console.