Ok, full interview from anonymous third party about Wii GPU.

[Entropy]

Maybe Nintendo should've gone with a 65nm Core Solo? Or some semi-custom 90 nm Dothan (i.e. a Celeron M but w/ speedstep enabled)? Either would surely destroy that special little 750CL, especially in performance per clock, if not power. But, it wouldn't have the backwards-compatibility-of-questionable-worth.

And of course, it wouldn't be as near-free as I bet 750CL most likely is. And that's what matters to Nintendo.

Dothan (at 90nm) was 84mm2 (five times the die size at the same process node) and had a TDP of 21W at 2GHz (a pure mobile product). Five times the size and five times the power dissipation at the same process node as the 750CL. Compared to the P-III in the benchmarks previously linked, the Dothan enjoys a better memory subsystem, but then again so does the Wii CPU compared to the 750s in the same link.

So you feel five times the die size, five times the power draw and lost backwards compatibility and the associated costs and delays in new tools would be a good trade for somewhat better integer performance? Insanity.

While it would have been very interesting to see what Nintendo could have achieved if they had kept the same cost and power envelope but had done a clean sheet design, adopting an x86 CPU would make no sense whatsoever. In spite of all the development dollars sunk into them, they suck, pure and simple, in low power applications. To achieve decent performance they carry a large die size overhead (cost) and they lack the custom functionality that might be useful in a gaming console.

 

[Eteric]

Hi guys, I'm new here, but I read these forums occasionally and I wanted to throw my thoughts in real quick.

I do think you're correct when you say that Nintendo spent very little on R&D for the Wii's hardware. Though it is indeed more powerful than the Gamecube, it lacks in comparison to the others.

BUT!...

What I think we're forgetting here is the Wii-mote it self. I seem to remember back before the Wii's release, when people were worried about the issue of sunlight effecting the Wii-mote, that Nintendo said they had tried many different types of motion sensing to find one that would less effected by the system's environment. The R&D costs of the Wii-mote alone are probably quite high.

As for the system itself, I owned a Gamecube, so I'm happy about it being backwards compatible. To many gamers, that's kind of a big deal these days. I think Nintendo went the direction they did because they wanted 100% backwards compatibility while talking up small amounts of space and energy (this is apparently a big deal to the Japanese).

We also have to take into account that if they had gone with higher end hardware, and the Wii failed, they would be in quite the spot.

One other thing before I head off. By the time the next generation rolls in, the Wii could probably be made near the size of a small video card for a PC, so Nintendo would be able to go with a better architecture, while keeping Wii and GC compatibility with shrunken Wii hardware (it'll probably be able to fit in a small case by then).

Anyway, that's my two cents.

 

[swaaye]

So you feel five times the die size, five times the power draw and lost backwards compatibility and the associated costs and delays in new tools would be a good trade for somewhat better integer performance? Insanity.

Wii is not a mobile platform. I really don't care at all how power efficient the thing is. We're not on battery here. And I don't really think Nintendo has it as high on their priority list as some folks here think.

Comparing a 7 yr old Tualatin to some brand newish revision of PPC was dumb, IMO. So, I brought in a more recent rendition of the P6 line. And I think Dothan and newer deliver a much better processor for PCs than that PPC, for sure. And I think you definitely underestimate how much faster they are than what amounts to a G3+. They especially seem to like games, in fact, if you look at all the reviews out there. In games, they embarrassed their bigger brother P4 and ran alongside the hugely more power-consuming Athlon 64 per-clock.

Considering Dothan/Yohah and gang use a whole 4 W at idle, I think they actually could fit the role somewhat well. The machine would be even easier to code for because it would have some major OOE resources and have a boatload of very fast cache. I don't think the 360 or PS3 have proved themselves to be undeniably superior to something like this. The only problem is that Intel doesn't want to be nice with its IP, unlike the PPC folks, and this causes manufacturing cost headaches down the road. At least it did for MS.

Nintendo went with what is undoubtedly an extremely cheap CPU, one that offers them backwards compatibility in the most direct way imaginable. I don't really find it to be impressive at all. I doubt it's even double the speed of what's in Cube. Maybe if it was in some subnote, but we don't see it in many systems do we? I wonder why? Perhaps because it's just too dog slow these days? Where does the chip get used, anyway?

I'm still waiting for the proven fact that "x86 sux" to actually mean something in the real world. Ever since people started going gaga over RISC and the PPC 601 arrived, that has been the rhetoric, and where has it gone?

 

[StefanS]

I do think you're correct when you say that Nintendo spent very little on R&D for the Wii's hardware. Though it is indeed more powerful than the Gamecube, it lacks in comparison to the others.

BUT!...

What I think we're forgetting here is the Wii-mote it self. I seem to remember back before the Wii's release, when people were worried about the issue of sunlight effecting the Wii-mote, that Nintendo said they had tried many different types of motion sensing to find one that would less effected by the system's environment. The R&D costs of the Wii-mote alone are probably quite high.

Ok, I am leaning myself out of the window here, but the R&D costs of the Wiimote are probably very miniscule. All the chips inside it are some standard chips with documentation readily available (in fact, that's the reason it only took a month to get the Wiimote working on the PC). They might have been tweaked ergonomics here and there and maybe circumvented a patent or two, but the rest is pretty standard.

 

[Capeta]

I agree with StefanS, Wiimote is basically a modern Power Glove...not much R&D spending needed.

 

[pc999]

What I think we're forgetting here is the Wii-mote it self. I seem to remember back before the Wii's release, when people were worried about the issue of sunlight effecting the Wii-mote, that Nintendo said they had tried many different types of motion sensing to find one that would less effected by the system's environment. The R&D costs of the Wii-mote alone are probably quite high.

As for the system itself, I owned a Gamecube, so I'm happy about it being backwards compatible. To many gamers, that's kind of a big deal these days. I think Nintendo went the direction they did because they wanted 100% backwards compatibility while talking up small amounts of space and energy (this is apparently a big deal to the Japanese).

We also have to take into account that if they had gone with higher end hardware, and the Wii failed, they would be in quite the spot.

Still a second core and 32-64MB more of Ram can bring significant beneficts and would keep full BC and would barelly increase the R&D/heat/size and even SW R&D wouldnt be that much increased to be well used (we are talking about a TnL architeture), the worst thing is that in 2 years they will be saving so litle that it is even sad, really sad...



Anyway there is any new hint in why is the GPU and CPU bigger than they are supossed to be

 

[bomlat]

I would imagine that the final documents list such details. No matter where this is managed it would seem pretty neccesary for a developer to know such basic info about the hardware he's working on.

first,the interwiev said that the doc was the same as the 2.1 sdk.And in the 2.1 sdk they have only a remark about a future wii gfx sdk.

 

[Entropy]

Wii is not a mobile platform. I really don't care at all how power efficient the thing is. We're not on battery here. And I don't really think Nintendo has it as high on their priority list as some folks here think.

Well, Nintendo has been extremely explicit in how power draw dictated the design, the end result can be compared to the other two new consoles in graphs that show an order of magnitude lower power draw ... I feel that your second sentence tell the whole story. YOU don't care how power efficient the Wii is. That's OK I guess, but your subsequent denial about its importance to either Nintendo or other consumers is not.

It would have been very, very easy and cheap for Nintendo to go with a higher power draw design. They didn't. Why didn't they, if it wasn't high on their priority list, just as they say?

That you can't understand the reasons why is something else entirely.

 

[bomlat]

Ok, I am leaning myself out of the window here, but the R&D costs of the Wiimote are probably very miniscule. All the chips inside it are some standard chips with documentation readily available (in fact, that's the reason it only took a month to get the Wiimote working on the PC). They might have been tweaked ergonomics here and there and maybe circumvented a patent or two, but the rest is pretty standard.

????think it again.The parts are standard,but the cost is not came from the parts.
that is came from the system itself.You have to found the right design,you have to found the right way to program this stuff.
The equipment to produce a chip is standard,the material is standard,so what make the price diference between 1 sqmm memory module and cpu?design cost.

 

[StefanS]

????think it again.The parts are standard,but the cost is not came from the parts.
that is came from the system itself.You have to found the right design,you have to found the right way to program this stuff.
The equipment to produce a chip is standard,the material is standard,so what make the price diference between 1 sqmm memory module and cpu?design cost.

Again the design is also fairly standard. Nothing out of the ordinary. 2 accelerometers and an IR cam. An EE student can create stuff like that after 2 semesters. Even in the initial stages of brainstorming using accelerometers was a given. All that was left to do was using a pointing system and for that there're many proven methods readily available. It's not wheel reinvented. If you think a fortune was spent on developing that you're deluding yourself.

EDIT: tried to clarify my point

 

[Shifty Geezer]

Again the design is also fairly standard. Nothing out of the ordinary. 2 accelerometers and an IR cam. An EE student can create stuff like that after 2 semesters. Even in the initial stages of brainstorming using accelerometers was a given. All that was left to do was using a pointing system and for that there're many proven methods readily available. It's not wheel reinvented. If you think a fortune was spent on developing that you're deluding yourself.

Definitely. I've said before, I actually find Nintendo's solution a bit simplistic. The Fusion (or whatever it's called) uses hypersonic sonar, in comparison. A triangulated signal was what I was expecting, and that solution has a lot going for it, other than cost...

 

[Eteric]

Ok, I am leaning myself out of the window here, but the R&D costs of the Wiimote are probably very miniscule. All the chips inside it are some standard chips with documentation readily available (in fact, that's the reason it only took a month to get the Wiimote working on the PC). They might have been tweaked ergonomics here and there and maybe circumvented a patent or two, but the rest is pretty standard.

Oh, I understand that. What I meant is that I think they've gone through a bunch of different motion sensing methods before getting to the final result. Surely doing so costs a bit of money. I seem to remember Iwata saying this (went through a bunch of different methods).

I could be wrong, though.

 

[Shifty Geezer]

Sure, but not millions. Imagine the components for each of 5 different prototypes cost $25,000 (and these things are supposed to be mass-consumer devices so can't cost too much) and you see it remains small-fry, even paying for engineers. Worst case you have a few engineers working for a couple of years on prototypes, and it's going to cost, I dunno, a few million dollars. That's not a cost you'll be needing to price into the Wii console, because it's peanuts. 50 cents on the console for the first 6 months (6 million sold) and it's covered. It's not going to push the Wii from a $200 console to a $250 console!

 

[bomlat]

The low cost can be true if you want to sell a motion sensing controller.The users complain, you get the experience, you improve and so on.
The wiimote was not developed by this way.Probably they did many test with thousans of testers,and later they scrap the desing.Later on,they had to test thousand times the final design,and they had to improve it to a very high level.
the dev period can be 2 year,and can involv at least 100 people.the cost of it have to be at least in the 30 million area.

On the other side, the N increase the price of the wii by 50$ because they could do that, and if they able to make 56$ instead of 8$ on every console,why not?Even the 250$ is not the optimal price.probably the 270-290 can be the good price.

 

[tongue_of_colicab]

Sure, but not millions. Imagine the components for each of 5 different prototypes cost $25,000 (and these things are supposed to be mass-consumer devices so can't cost too much) and you see it remains small-fry, even paying for engineers. Worst case you have a few engineers working for a couple of years on prototypes, and it's going to cost, I dunno, a few million dollars. That's not a cost you'll be needing to price into the Wii console, because it's peanuts. 50 cents on the console for the first 6 months (6 million sold) and it's covered. It's not going to push the Wii from a $200 console to a $250 console!

Isnt the bigges amount of cost related to labour? I dont know how much nintendo spend to make the controller, I dont want to mix in that discussion either but I figured paying the people who worked for over a year on it is probably alot more expensive than making the prototypes.

 

[Eteric]

No idea, it's unfortunate for us that Nintendo is such a secretive company.

I also recall Iwata stating that Nintendo was working on a powerful console like the 360 and PS3, but scrapped it for the Wii. I wonder if they paid a whole lot of money to have that console developed? Unfortunate we don't know how long they had worked on it, so it's impossible to tell.

My biggest interest is seeing how far developers can push the Wii. I've always had some kind of weird interest in the weaker consoles (PS2, DS, etc) because sometimes developers do amazing things with so little. Can't wait to see what they do in the future.

Anyway, so no one knows what that extra space in CPU (or was it GPU?) is for? I'm not a hardware person by any means, but couldn't they look inside of it?...

 

[tongue_of_colicab]

If I remember correctly what Iwata said was that they worked on a powerfull next gen console for a not so long time before they switched to what now is the Wii. I remeber him saying that he couldnt really press his mark on the DS because it was already to far in development when he took things over from Yamaguchi but he could make the Wii like he wanted though they already did some work on it wich they apparently ditched.

 

[swaaye]

Well, Nintendo has been extremely explicit in how power draw dictated the design, the end result can be compared to the other two new consoles in graphs that show an order of magnitude lower power draw ... I feel that your second sentence tell the whole story. YOU don't care how power efficient the Wii is. That's OK I guess, but your subsequent denial about its importance to either Nintendo or other consumers is not.

It would have been very, very easy and cheap for Nintendo to go with a higher power draw design. They didn't. Why didn't they, if it wasn't high on their priority list, just as they say?

That you can't understand the reasons why is something else entirely.

Why didn't they? Because, to me at least, it's fairly obvious (I think) that they wanted to have a machine on which they could recycle their dev tools (and even their game engines) from last generation. A new angle on console cost cutting, for sure.

And throw in the now-apparently-ubiquitous backwards compatibility bullet. I'm not sure where this was on the priority list, but I think it could very likely just be a convenient aftershock of their reusing Cube tech.

All of this conveniently coincides with a low-power design because the hardware they wanted could be built mighty small today. 180 nm hardware reduced to 90 nm means some serious power gains, especially considering Cube was already power frugal. At least they changed the GPU a teensy bit.

How very exciting. If the Wii didn't have it's semi-gimmicky controller (I put it at 25% gimmick), it sure wouldn't be thrilling at all from a technological perspective. (I do own a Wii, btw.)

I'd be less bitchy if they'd at least made some new, interesting hardware. I don't see a reason something built on 90 nm can't be an absolute monster for 480p and yet still be low-power, but Wii certainly isn't such a beast. Hell, at least stick in some more RAM!

 

[tongue_of_colicab]

They just wanted something they could sell dirt cheap in needed and something wich would make it real easy for devs to work on and they needed something else to make people buy their console if they'd go for low tech hardware.

Doesnt it make sense if you think about it? if they'd make another ''expensive'' powerhouse, it wouldnt sell just as with n64 and gc. They needed something that would pull consumers and devs in. Using upgrades of your old hardware and software seem the fastest, best and cheapest option to me. They trew in the controller to make the differance between other consoles.

Yes they could have trown in better hardware, but that would make it more expensive (aka less profit) and would make things harder as devs have to work with new hardware instead of the old one, even if they didnt use the GC hardware, atleast there is plenty of info on coding already wich makes things easier.

Atleast from a nintendo pov it makes sense I think.

 

[MDX]

I recall Nintendo stated they wanted to make a console to be sold for $100, but the technology they put in the machine doubled the price.
I also recall that the disc drive was expensive to design too.

At any rate, people who are saying Nintendo should have done this or that are talking out of their butt. According to sales, Nintendo has done everything right this time around. And they havent even released their big three titles yet! Im curious what under the hood of the machine, but I dont think they needed to improve on anything.

In the end it comes down to if you enjoy the games that will be coming out on the console. Japanese developers are already shifting focus to the Wii away from the PS3, EA is supporting it like crazy, and we are going to probably see a slew of titles from small independent developers. Its all about the games. And in two years time I predict Wii will have the most interesting and engaging games compared to the other two consoles.