memba itz 729MHz (1.1 GHz) + 243 MHz ssj12, so it is double the GC...

Woah, 16 pipes? Then the Wii packs quite a punch.

Anyway, anyone knows the original FSB of the GC? And the multiplier?

I still believe the Wii has something in the likes of a Radeon X1000 or something like that, but maybe we will never know. 

As much as the architectures are very similar you can't directly compare performance simply due to the clock rate ...

The manufacturing process of the Gamecube's processors was 180nm whilst the Wii's processors use the 90nm process; if the Wii used the same processors as the Gamecube the die size of the Hollywood and Broadway processors would be 1/4 the size of the Flipper and Gekko yet verified reports have the Hollywood and Brodway processor's die size at a little more than 1/2 that of the Flipper and Gekko.

The simplest (and most logical) explaination is that Nintendo has increased the number of transistors by adding cache and instructions to the Gekko to produce the Broadway processor whilst increasing the number of texture and lighting pipelines on the Flipper to produce the Hollywood processor.

The end result would be that added cache would increase the efficiency of the Broadway processor by lowering the number of times it has to go to general memory; the added texture and lighting pipelines on the Hollywood processor would enable it to process more texture data and lighting effects per frame.

You do some strange math.

Gamecube CPU = 485 MHz, GPU = 162 MHz.  Wii CPU = 729 MHz, GPU = 243 MHz (as far as we know).  Both are 50% faster clock speeds, you can't just add those together and get "double".  Of course, there are also other factors involved, as HappySquirrel pointed out.

CrashMan, you said you found the information you were looking for, care to link us to it?  :)  I think what you read said 16 stages, not pipelines.  The Flipper had 4 pipelines, and based on the die size of the Hollywood, it is believed that it may have 8 pipelines.  But if you have information that says otherwise, please share.

Analyzing the die size is kind of an inexact science in and of itself.  The Gamecube's GPU die was also the system chipset, including I/O, a DSP sound processor, and a memory controller.  Hollywood actually has two dies, one of which has the GPU, I/O, and memory controller, and the other containing the DSP and the 24 MB of internal memory (which corresponds to the Gamecube's 24 MB of system memory).

 That was mean Dolla Dolla. You stole my joke.

Hey Entroper, about the dies on Hollywood. One is supposed to be called Vegas and the other Napa. How do we know the names? They have the names printed on them or something? I haven't opened my Wii.

How do they know what does what on the dies?

Also, you happen to know if someone has die-shots of any of the chips? Something like this:

One of the dies on Hollywood happens to be the exact same size as a RV515... aka the Radeon X1300. I know it may be a coincidence, but good to know ain't it?

Thanks for explaining the PixelShader issue Entroper.  The Wii is even more amazing than I thought :P.

As for the CPU discussion, the Wii's CPU is a PPC 750CL. It is not a 750GX since the GX is manufactured using 130nm not the 90nm the CL and Broadway use. The GX is also 52mm^2 which is much to large to be used as a base for the 19mm^2 Broadway (even accounting for using 90nm). Also, according to Wiki at least, the CL is just a lower power consuming CXe with various minor additions. This is IBM's datasheet for the CL http://www-01.ibm.com/chips/techlib/techlib.nsf/techdocs/2F33B5691BBB8769872571D10065F7D5/$file/750cldd2x_ds_v2.4_pub_29May2007.pdf
Broadway reportedly uses 20% less power than Gekko so the Broadway must be below the 733MHz speed since that one uses the same power are Gekko did. I know 729 isn't much below 733 but as it says in the notes, different voltages, bus speeds, and mulitpliers will cause different results. Considering the 1 GHz model uses 10.5 W I very much suspect Broadway's speed is below 733 MHz and not above it (I'd be willing to bet it's probably somewhere around 729Mhz).

Interestingly, Gekko was the same 43mm^2 as a standard 750CXe but Broadway is 18.9mm^2 which is somewhat larger than the standard 750CL's 15.9mm^2. I wonder what the extra 3mm^2 is used for, 256KB of cache, extra FPU's (I don't know if that would do any good), a second hidden core (I'm just kidding on that one)?

EDIT: Actually, ignore most of my post.  Reading through this thread http://forum.beyond3d.com/showthread.php?t=34269 I'm not so sure about about the 750CL.  Knock off half the 1MB cache of a 750GX and scale down to 90nm from 130nm and you have about 17-21mm^2.  Also there is some confusion about what "20% less power means".  Is it overall, average, over Gekko at the same clock speed, or on a MHz/W basis.  Why everyone would say it's a 750CL if it isn't is my main problem with the GX theory.

Yes, the clock can be changed through software... firmware to be more exact. To whoever has a Wii, you can notice that while playing games, the console doesn't gets too hot. This gives us headroom for a good ol' overclock.

Also, the console can throttle back to a slower speed when in idle mode, like the AMD Cool n' Quiet tech... but this all depends if the CPU has such capability.