Nintendo Gamecube
Nintendo's next-generation console is so vastly different in design and execution from PlayStation 2 that it's an incredibly difficult task to try to compare the two consoles. Gamecube is powered by a 405MHZ PowerPC-based chipset called "Gekko" that features several extensions optimized to process OpenGL code -- an API familiar to PC developers and used commonly for games like Quake and Unreal. On top of that the Gekko features 256k of L2 [Level 2] cache, which is extremely useful when processing code quickly and efficiently.

The console also features a 202.5MHZ ArtX developed graphics chip codenamed "Flipper" with 3MBs of embedded 1T-SRAM and onboard geometry and lighting engine. It also boasts hardwired S3 texture compression for on-the-fly decompression with no hit on hardware.

Gamecube includes a total of 40MBs additional RAM, 24MBs of which is 1T-SRAM -- a blazingly fast solution for the main RAM, and 16MBs of which is PC-100 DRAM used primarily for animation and audio purposes.

Whereas PlayStation 2's CPU and two Vector Units split up the tasks of various graphic procedures, like transformation and lighting, for example, all of this is handled singularly by Gamecube's Flipper chip, which also decompresses textures at a 6:1 ratio. PS2 has no hardware texture compression and seeing as how it only features 4MBs of Embedded DRAM on its graphic synthesizer, developers would need to compress textures in software, which in turn means a significant hit on the Emotion Engine CPU.

Michael Lamb, CEO of Left Field Productions, offers: "One of the bottle necks the PS2 developers I talk to seems to be the limited size of video memory. This will be less of a problem on the Nintendo Gamecube because of the speed of the memory and S3 compression resulting in smaller textures."

Furthermore, Gamecube renders up to eight effects layers to a polygon in a single pass, whereas the PS2 features a multi-pass rendering system. So, for example, Gamecube developers can effectively start with the base geometry (1), add a bump-map to it (3), add a dirt map (4), add a gloss map (5), add a reflection map (6), add a radiosity light map (7) and an effects layer of their choice (8) -- all in a single pass. By contrast, PS2 developers would have to re-render the polygon itself for every pass meaning eight times the work to get the same effect. So essentially PS2 has to render 1,000 polygons eight times over whereas Gamecube only has to render 1,000 polygons once for the same effect.

"With PS2 you have to write the code to do a lot of that stuff. But that's the trick behind it," says Gregory P. Zeschuk, president and CEO of Bioware, which created MDK2 for Dreamcast and is now bringing the title to PS2. "On the one hand the downside is that it's not done for you. On the upside, if it's not done for you then you can do it yourself, and potentially better -- more specialized to what your specific application is. So it's really a double-edged sword. That's the mystery of the system."

Still though, to have access to multi-texturing features from the beginning versus having to fight through code for the process is a definite advantage for developers hoping to create timely software. "Yes, it's very nice that Nintendo Gamecube can do eight layers in one pass. It's all set up for you. Believe me, I would have loved it," says Jason Rubin of Naughty Dog. "But we'll have eight-pass techniques done on PlayStation 2 by the time we're done with our first game. So it's automatically done on Gamecube, but we've got the same techniques. Yes, they take more time and yes they take more CPU power, but here we go again -- [PS2] is a more powerful CPU so who cares? In the end you break even."

But is PS2's CPU with Vector Units really more powerful than Gamecube's? We went to Julian Eggebrecht, president of Factor 5, a developer currently underway with Gamecube projects for a different opinion. "Gamecube's Gekko is the most powerful general-purpose CPU ever in a console. The PowerPC alone is so much better and faster structurally that Gekko not only is much, much faster than the PS2's main CPU but every bit as fast as a 733 MHz Pentium," rebukes Eggebrecht. "Don't forget how extremely powerful the 256K second level cache in Gekko makes Gamecube. The size of a CPU's second level cache determines how fast general game-code runs. The PS2 has a tiny 16K of second level cache, and even the X-Box only has 128K."

In terms of how it performs against PS2's Vector Units, Eggebrecht offers the following: "Gekko is not just a plain PowerPC CPU, it has special commands just for games and has special modes making it possible to run hand-written assembler code very, very fast. We did experiments with particles on Gamecube after the N64, and as opposed to the two weeks it took to get the particle system running and optimized on the vector unit it only took two days on Gamecube.

"Based on our calculations, Gamecube's CPU has all the power PS2's CPU and VU0 have combined and then some. Everything you do on VU1 on the PS2, Gamecube has in hardware in 'Flipper,' the graphics chip. And believe me, when it comes to fill-rate and polygon counts, Gamecube is every bit as good as PS2. Nintendo has chosen not to throw around numbers, but we know those numbers and as opposed to what Naughty Dog might think they are not necessarily lower than PS2's numbers."

There is much more to factor in, though, than just CPU speed and graphics chip architecture. Eggebrecht points to Gamecube's RAM set up, as well as its texture and lighting capabilities, as further proof of its power. "Another important area is the RAM," he says. "RAM is where your game and data are living in and being moved through and what really determines the speed and power of a system to a large degree. Anybody who ever programmed a N64 knows that Rambus RAM is especially problematic for game purposes when CPUs have to randomly access tiny chunks of data in a very fast manner. The PS2 has that same Rambus RAM. The more you use the CPU to randomly move chunks of data, the more speed problems you get.

"Gamecube has a completely new technology, MoSys RAM. The speed of it is unbelievable, something we noticed when we started throwing more and more data and game logic at the machine. RAM-speed simply is not an issue for Gamecube. That is a first among the next-gen consoles and a major breakthrough."

While which system ultimately pushes more polygon counts is debatable, there is no arguments about Gamecube's superiority in what can be done visually with those polygons. Naughty Dog's Jason Rubin concedes: "The Gamecube is going to have a great renderer, and probably will put out almost as many polygons as the PS2, but they will be prettier polygons. Its problem is going to be, truthfully, that its CPU is weaker so it won't be able to do as many cool things with those polygons."

Meanwhile Eggebrecht comments: "The area that Gamecube is similar to PS2 but just plain kills it is in textures and hardware lights. There is simply no comparison. So you get similar polygon counts on GC as you have on PS2, but with these incredible amounts of texture effects on top. Just imagine a textured, bump-mapped, reflecting and accurately glossy and dirty car with specular highlights thrown in - that's just one pass for Gamecube but many for PS2. And of course there is the Gamecube texture decompression thrown in for free, something the PS2 doesn't have at all."