WiiGeePeeYou (Hollywood) what IS it ?

[fearsomepirate]

Hmm, those graphics look fairly on par (imo, it's hard to compare) to the old red steel and rayman graphics. A bit worse perhaps, but beyond anything on the wii so far.

I've seen a higher-quality vid of the pokemon getting blasted with green fire, and it's truly an impressive effect. Not only is the distortion pretty high-res and fills up a good chunk of the screen (never saw anything on that scale with Cube), but it also layers on top of itself.

The effect's definitely an indirect-texturing sort of thing, since it's in the same genre as stuff we've seen on Cube, but IMO it's a lot more than a mere 50% overclock would indicate. Maybe they added a pipe or two, or maybe the TEV's been beefed up to do more effects per pass.

 

[Rolf N]

Wait, what's going on here? Hollywood's size is actually bigger than Broadway? And the die shrinkage don't jive with the 90 nm process?

It pretty much does. The CPU size is in line with what theafu called. The expectation going in this thread here that pad limitation will result in losing some fixed percentage of the die per shrink, irrespective of the absolute size of the die and the number of pins is just ... misguided. Unwise. It fits the prediction well.

Hollywood looks like it might be a doubled Flipper with the same amount of edram, which would mean forced anamorphic widescreen. Awful.

 

[nonamer]

You can send the winnings anyway you like, cheque, postal order ect, but I'd prefer PayPal

I wasn't betting, only pointing out it's much more likely not to be there rather than being there.

That said, there really is no point of putting 1T-SRAM on package. The reason I can think of is that 1T-SRAM is really hot running and thus needs a heat spreader.

 

[Shogmaster]

Thank goodness someone finally peeked under the heat spreader, but now I'm more confused then ever.

The RAM is on a seperate die, but in the same package. Why? Why go through the troble if you can't use it like 360's EDRAM? Isn't it just more costly without having any performance gains?

At least we know for sure that the 24MB of 1T-SRAM is still there.

 

[thejeek]

Isn't it just more costly without having any performance gains?

maybe the bus is wider? maybe the 24MB 1-T can be used as backbuffer directly? The second seems to me to be pretty much a requirement for widescreen PAL support - otherwise there's too many pixels to fit in a 2MB framebuffer (I think regular PAL is already slightly too big for the Cube frame buffer with square pixels, widescreen would require seriously stretched pixels horizontally)

 

[fearsomepirate]

Not great physics, but going from gamecube where not a single game used physics to any notable extent

What qualifies as a "notable extent?" There were games with ragdoll for dead bodies, the Monkeyball games, other games using Havok or something for boxes and such, etc.

 

[pc999]

Wait, what's going on here? Hollywood's size is actually bigger than Broadway?

It always is.

Is it even a good idea to have a GPU with a pretty good jump in abilities, with a CPU with very low jump (assuming this is what everyone's talking about)?

It can be if those features a) offload the CPU (eg VS, physics ...) or 2) those features arent dependent of the CPU (eg unlike a new TnL engine).

maybe the bus is wider? maybe the 24MB 1-T can be used as backbuffer directly? The second seems to me to be pretty much a requirement for widescreen PAL support - otherwise there's too many pixels to fit in a 2MB framebuffer (I think regular PAL is already slightly too big for the Cube frame buffer with square pixels, widescreen would require seriously stretched pixels horizontally)

Didnt the 2MB framebuffer do tiles?

What qualifies as a "notable extent?" There were games with ragdoll for dead bodies, the Monkeyball games, other games using Havok or something for boxes and such, etc.

I think he mean from those boxes to elebits physics, althought we really dont know what Gekko can do in terms of physics because only a few multiplatform PS2 from games did anything, (besides the water physics in Mario Sunshine, that seems very good and at 60 FPS).

Althought if he really mean that game it is still hard to get a good opinion once that the game dont seems to stress anythung else (althought they refer the AI as a strong one too)

 

[Teasy]

It pretty much does. The CPU size is in line with what theafu called. The expectation going in this thread here that pad limitation will result in losing some fixed percentage of the die per shrink, irrespective of the absolute size of the die and the number of pins is just ... misguided. Unwise. It fits the prediction well.

Nobody is trying to claim that there is an exact percentage of die space that will always be wasted because of pad limitations. What I an others were trying to do is be conservative by assuming that a decent amount of the added space was not taken up by added logic or memory transistors. So that we could discuss more realitically what could be achieved in the apparent extra die space.

 

[Squeak]

Hollywood looks like it might be a doubled Flipper with the same amount of edram, which would mean forced anamorphic widescreen. Awful.

One megabyte with 24bits per pixel is enough for something like 720x480. Not too shappy.
Although full 853 horizontal resolution would be nice and it would only require ~400Kb for a full framebuffer.

 

[Teasy]

I wasn't betting, only pointing out it's much more likely not to be there rather than being there.

Only joking, no point in arguing now after all. I explained at the time why I felt it was more likely that it was there and, well at the end of the day it is so..

 

[Fox5]

What qualifies as a "notable extent?" There were games with ragdoll for dead bodies, the Monkeyball games, other games using Havok or something for boxes and such, etc.

I don't particularly remember physics in the monkey ball games besides the monkey bouncing, and the wii sports physics seem better than most havok/ragdoll stuff. They're running at close to real world speed, rather than the exaggerated slowness ragdolls usually have, and the movement of the pins in bowling seems pretty good. They'll fly off in the correct direction when hit, spin, and tilt. They appear to correctly model real pins, outside of being weightless.

(besides the water physics in Mario Sunshine, that seems very good and at 60 FPS).

Mario Sunshine was 30fps.

 

[pc999]

Mario Sunshine was 30fps.

Are you sure, anyway I will tak your word for it as I dont remeber very well.

 

[RocketMBA]

Are you sure, anyway I will tak your word for it as I dont remeber very well.

I seem to remember it as 30 as well... But there's only one way to find out! Give me a second and on goes the GameCube...

Yeah, 30 frames a second.

 

[Rolf N]

Nobody is trying to claim that there is an exact percentage of die space that will always be wasted because of pad limitations. What I an others were trying to do is be conservative by assuming that a decent amount of the added space was not taken up by added logic or memory transistors. So that we could discuss more realitically what could be achieved in the apparent extra die space.

Sorry for being so nondescript, I was awfully tired when I wrote that. No idea what made me post at that hour.
I was thinking about the percentage derived from the Athlon 64 shrink, and still don't think that chip is in the correct ballpark for a comparison. The pads will "waste" a larger percentage of the die on Broadway than they do on a current A64. Anyhow that wasn't relevant anyway, as we were working with measurements already at that point.

I still don't believe we have much "extra die space" on that picture. IBM specs 15.9mm² for the chip theafu called. While the measurement says we're at 18.9mm², that's less than half a mm of wiggle room in each dimension. Maybe they are just wrong. Perpetual rule of the Wii launch:
Don't get your hardware hopes up without a very strong reason.

One megabyte with 24bits per pixel is enough for something like 720x480. Not too shappy.
Although full 853 horizontal resolution would be nice and it would only require ~400Kb for a full framebuffer.

More of too-tired posting on my part
You're forgetting the z-buffer in your calculations. But no worries, what I wrote actually doesn't make sense at all. If Hollywood is straight 2xFlipper, the edram would be doubled too. 4MB of framebuffer memory would be more than enough for proper widescreen.

The large Hollywood die is very interesting, but not for that reason

 

[RocketMBA]

On the subject of proper widescreen, I read somewhere (the Wii manual?) that QuickTime movies and JPEGs will output at a maximum of 848x480. I suppose it'd be silly to say something like that if that wasn't the external resolution as well.

 

[Teasy]

Sorry for being so nondescript, I was awfully tired when I wrote that. No idea what made me post at that hour.
I was thinking about the percentage derived from the Athlon 64 shrink, and still don't think that chip is in the correct ballpark for a comparison. The pads will "waste" a larger percentage of the die on Broadway than they do on a current A64. Anyhow that wasn't relevant anyway, as we were working with measurements already at that point.

In my post I used the example of the Athlon64 as my best case scenario (25% I believe it was) and 50% as my worst case. I suppose it would be much easier if we knew 100% for sure that 750CL was purely Gecko on a 90nm process. Because then there would be no need for guessing. There would only be one question, are the measurments correct.

I still don't believe we have much "extra die space" on that picture. IBM specs 15.9mm² for the chip theafu called. While the measurement says we're at 18.9mm², that's less than half a mm of wiggle room in each dimension. Maybe they are just wrong. Perpetual rule of the Wii launch:
Don't get your hardware hopes up without a very strong reason.

Of course its a possibility that they could be wrong. But I don't think there's any reason to assume it and rule out all speculation on what possible extra die space could hold Though I admit its less obvious then in the case of the GPU where it seems certain there are upgrades over Flipper.

 

[Megadrive1988]

Mario Sunshine was originally presented at 60fps. from SpaceWorld 2001, up until E3 2002. there were videos that clearly showed it at 60fps, and the ultimate proof was the footage that aired on CNN during E3 2002, the broadcast/cable television footage clearly showed the game running at 60fps, without a doubt. but the next batch of videos on IGN showed they dropped it to 30fps, and that was what the final game was, 30fps.

I'm told on first-hand authority that it's a solid 60.

Zelda Twilight Princess on Wii is a solid 30fps - it is absolutely *not* 60fps - same as the Gamecube version, 30fps.

 

[Megadrive1988]

And why there appears to be so little benefit from it - with 1.5x clock and nearly twice as many transistors as Flipper, shouldn't the visual improvement from the Cube be more obvious?

because most of the Wii games we'vee seen were developed on Gamecube or overclocked Gamecube dev-kits without the final Hollywood GPU. thus, they look like Gamecube or Gamecube+ games.

 

[Fox5]

Mario Sunshine was originally presented at 60fps. from SpaceWorld 2001, up until E3 2002. there were videos that clearly showed it at 60fps, and the ultimate proof was the footage that aired on CNN during E3 2002, the broadcast/cable television footage clearly showed the game running at 60fps, without a doubt. but the next batch of videos on IGN showed they dropped it to 30fps, and that was what the final game was, 30fps.




Zelda Twilight Princess on Wii is a solid 30fps - it is absolutely *not* 60fps - same as the Gamecube version, 30fps.

Sad, if Wii really was a dual flipper at 243mhz, 60fps widescreen should have been easy. Judging by how games aren't easily exploiting a 2-3x increase in power over the cube, I'd say it's more likely additional features/programmability has been added than a doubling of pipelines or the TEV.

 

[Megadrive1988]

Fox5 yeah I agree - although Miyamoto said there was some discussion about taking advantage of Wii's extra overall power and beefing up the graphics. Miyamoto discouraged that idea, saying that the team should focus on finishing the game, making everything right in terms of controls and everything going into both versions of the game.

I'm sure Nintendo is saving Wii's extra performance, as far as Zelda, for the next major game.