WiiGeePeeYou (Hollywood) what IS it ?

[bomlat]

EA patent

So, let see what is the official oppinion of the EA:

The GPU is user programmable. (This is not strictly true in the case of the Nintendo GameCube, but a rich set of predefined computational elements is available.)

 

[Teasy]

It looks like a refined Cube-level game. It looks like what you'd get with the legendary Cube + 50%.

The depth of field blurring hides the background detail/texturing well. The bike and character are fairly detailed, but the textures are still a bit iffy. It looks ok though. Cars look decent. I don't think it has anything on Burnout Revenge on Xbox, say, really. The game looks like it may have some sort of shadowing going on, in some shots.

This is an open ended city based game we're talking about here. You have to take that into account and can't just compare it to any game. Look at games like True Crime on GC/XBox, or GTA on PS2/XBox (games in a very similar genres) and compare. Those screen shots compare more with Saints Row then anything on GC or XBox AFAICS.

True Crim New York (GC):

Driver

Again this is a case of if those images are truly representative of the final game (the aliasing does seem to suggest its at least in engine). But if so IMO its a significant leap over anything in the same kind of genre seen on GC or XBox.

 

[The_legend_of_drtre]

Take a look at this old artcile from Anandtech

http://www.anandtech.com/showdoc.aspx?i=1566&p=13

"The GameCube wins in terms of GPU efficiency courtesy of the embedded 1T-SRAM from MoSys. However the use of a fixed function T&L pipeline is a bit of a turn off for the GPU. Again this is another situation in which it would have been beneficial to have ATI's input into the design of the product before it was finalized. It is a shame that ATI acquired ArtX after the design was already completed otherwise we might have seen a programmable T&L pipeline instead. "

It states that the reason why they believe Flipper had a fixed T&L is because ATI was not directly involved in the creation of the GPU. Artx had the chip down before ATI bought them out. This time ATI handled the GPU enternally. Maybe it's not fixed this time.

 

[Teasy]

I doubt that would have made any difference at the time. When Flipper was developed fixed function T&L was all there was in the GPU space, at ArtX, Nvidia or ATI. Obviously your point is that ATI may add it now, but to be honest ArtX could have added it in this day and age, its just a question of wether Nintendo asked for it. At the moment all the info suggests they didn't. Though perhaps this is another case of "Flipper has no programmable shader". In other words people confusing Hollywoods abillities because its T&L is not a standard vertex shader, (like TEV wasn't a standard pixel shader). Perhaps Hollywood does have flexible/programmable T&L but in a slightly different way to a traditional vertex shader. Or perhaps Nintendo just didn't see the need for any programmable T&L so Hollywood just doesn't have it at all..

Still it seems strange that even Nintendo DS has a programmable T&L engine and Wii doesn't....

 

[pc999]

Or perhaps Nintendo just didn't see the need for any programmable T&L so Hollywood just doesn't have it at all..

In the Goto article they said that programable shaders would lead to the games being harder to dev( , would it be that harder ), also it would increase the heat/power requeriments. BC also would be problematic (but IIRC, it isnt the main reason).

Still it seems strange that even Nintendo DS has a programmable T&L engine and Wii doesn't....

This put the power/heat requirements justification in a hard position .

BTW I didnt know that, can you say how much flexible they are or share more info, please.

 

[tongue_of_colicab]

Still it seems strange that even Nintendo DS has a programmable T&L engine and Wii doesn't....

I didnt know that. I asked this before at a dutch tech forum because I thought the DS had some kind of gpu in it but they told me everything was done by the ARM's so no T&L.

 

[swaaye]

It can push way more polys than an N64, so that's a big clue right there. (that it has HW T&L)

 

[DeadlyNinja]

Can it? I keep hearing conflicting reports about the amount of polygons the Ds and N64 can push. The original Nintendo Power and Next Generation magazine has the N64 listed at 100,000, while other online sources claim 150,000. The DS's number has always been 120,000 though. What is the actual number for the N64?

 

[swaaye]

It was around 150K for N64, unless the devs got creative with the microcode. But I've heard DS can tear through more like 2 million /s but may be limited in other ways that bring that number down.

 

[DeadlyNinja]

Wow, 2 million, are you serious? Even if there is a limit, a 50% decrease is still 1 million! Then what's the deal with 120,000 number that Nintendo gave out? Maybe they mean a single frame is composed or 120,000? The Ds is set to run at 60fps. That doesn't make sense though, since that would equal over 7 million polygons.

 

[Squeak]

It's something about the polygons being buffered, either for sorting or other reason and that buffer has a size limit.

 

[swaaye]

Ye Olde DS GPU thread:
http://www.beyond3d.com/forum/showthread.php?t=31794&highlight=geometry

Don't know how accurate this info is:

Maximum 4 million vertex per second geometric transformation
30 million pixels per second maximum fill rate
120,000 polygons per second maximum

That thread is very informative though. DS is definitely faster than N64. Just check out how they enhanced Mario Kart DS. Of course, could be better understood 3D coding techniques too.

 

[Turok]

Again this is a case of if those images are truly representative of the final game (the aliasing does seem to suggest its at least in engine). But if so IMO its a significant leap over anything in the same kind of genre seen on GC or XBox.

Setting aside the fact that the True Crime screens are from the PS2 version, they're bad captures.

http://xboxmedia.ign.com/xbox/image...rue-crime-new-york-city-20051107010636884.jpg

http://xboxmedia.ign.com/xbox/image...rue-crime-new-york-city-20051107010637337.jpg

I wouldnt call Parallel Lines "a significant leap" over other open ended games.

 

[Corwin_B]

Wow, 2 million, are you serious? Even if there is a limit, a 50% decrease is still 1 million! Then what's the deal with 120,000 number that Nintendo gave out? Maybe they mean a single frame is composed or 120,000? The Ds is set to run at 60fps. That doesn't make sense though, since that would equal over 7 million polygons.

IIRC, 4 millions ploys/sec is the theorical throughoutput for the DS GPU. The 120k polys/sec is a display limit, but it's 120k polys actually displayed (once clipping, backface culling... operations have been done).

 

[Shifty Geezer]

DS is definitely faster than N64. Just check out how they enhanced Mario Kart DS.

As empirical evidence, that's not an accurate comparison. Certainly DS renders to a smaller screen - 256x192. Weren't N64 games mostly at 320x256? That's 1.6x the drawing requirements there. Targetting the lower res, they could get more into the DS on the same hardware. And as you say, 3D techniques could be enhanced, plus other design choices. For whether DS is more powerful than N64 or not, the achievable specs need to be considered, or a glut of games noticeably superior on the handheld.

 

[fearsomepirate]

As empirical evidence, that's not an accurate comparison. Certainly DS renders to a smaller screen - 256x192. Weren't N64 games mostly at 320x256?

This is only relevant if the polygon draw-rate is fillrate-bound, not transform-bound. In general, we've been seeing more geometry, better texturing, and better framerates on DS. I doubt that's all due to a resolution change. Somebody already gave us the DS's max theoretical transform speed, and it's higher than N64's. Even without the bilinear, I think N64 maxed around a half-million polygons.

 

[djdev]

How do you know that though? Just because any new engine couldn't be Unreal Engine 3 in all its glory doesn't mean it couldn't be better then the current Unreal Engine 2 on GC/Wii

Incidentally I was going to post this in the Unreal Engine 3 on Wii thread, but its closed so I'll post it here instead:



So it looks like we might end up seeing once and for all wether this can be done and what advantages it may bring.

I really would like to see what that licensee (Gearbox???) can do with it. We know the Wii can do some normal mapping (whether it's in hardware or software, we don't know). I wonder if they get it to work well, will Epic then try to pursue releasing a version of it.

 

[Shifty Geezer]

This is only relevant if the polygon draw-rate is fillrate-bound, not transform-bound.

Oh, definitely. It was just one example of what's different. Comparing different screenshots isn't necessarily a good way to determine hardware power though...
[/quote]In general, we've been seeing more geometry, better texturing, and better framerates on DS. I doubt that's all due to a resolution change. Somebody already gave us the DS's max theoretical transform speed, and it's higher than N64's. Even without the bilinear, I think N64 maxed around a half-million polygons.[/quote]Going by the numbers in this thread, yes. 100' for N64, 150k for DS. However, checking Wikipedia, that suggests with different microcode, N64 could get 500k polys. Certain specifics need to be considered to turn specs into actual hardware performance, such as quality of polygons drawn.

This is what makes detective work from screen shots so interesting! A nicely filled out hardware document would put an end to our delicious 3000 ost thread, but until then, we have to try and gauge from screenshots and chip labels what the hardware is capable of, and that means taking into consideration all the differences of the screenshots and what might account for them.

 

[Dr_Lazarus]

I'd really be happy at this point just to see a pic of Hollywood's die. I guess the person who took the pic minus the heat spreader off wasn't brave enough to try and remove the thermal paste. I just wish I could get a hold of a broken one... Unfortunately my curiosity isn't quite strong enough to make me sacrifice mine.

 

[Fox5]

Oh, definitely. It was just one example of what's different. Comparing different screenshots isn't necessarily a good way to determine hardware power though...

In general, we've been seeing more geometry, better texturing, and better framerates on DS. I doubt that's all due to a resolution change. Somebody already gave us the DS's max theoretical transform speed, and it's higher than N64's. Even without the bilinear, I think N64 maxed around a half-million polygons.[/quote]Going by the numbers in this thread, yes. 100' for N64, 150k for DS. However, checking Wikipedia, that suggests with different microcode, N64 could get 500k polys. Certain specifics need to be considered to turn specs into actual hardware performance, such as quality of polygons drawn.

This is what makes detective work from screen shots so interesting! A nicely filled out hardware document would put an end to our delicious 3000 ost thread, but until then, we have to try and gauge from screenshots and chip labels what the hardware is capable of, and that means taking into consideration all the differences of the screenshots and what might account for them.[/QUOTE]

DS could even have an inferior gpu to the n64 in performance, but still perform better solely due to its faster memory.