Now that this thread has slowed down, I'm going to post something I sent to IGN's mailbox:
Dear Matt,
While I do believe you have the best information available, simple logic would dictate that the revolution be more than twice as powerful as an XBox. I'm not expecting it to be in the same league as the PS3 and 360, and don't really care. A couple of weeks ago I dug F-Zero GX out and had another spin on it. I live in Japan and was playing through the D-terminal cable on a standard definition set (though with scanline compression to 16:9) and thinking, damn this still blows me away. At that point I realized I agreed with Nintendo's strategy. I don't need another fidelity jump the likes of N64 to Gamecube, that won't be enough to make me buy a new system. That being said, I expect more than 2 times a Gamecube. 3 times a Gamecube or higher and I'll be happy. By my reasoning, that should be very simple for Nintendo to reach without architecturally changing a single aspect of the gamecube... through a process shrink.
Gekko has roughly 20 million transistors and flipper around 26 million (or so google tells me). Gamecube never had a process shrink so those chips are still produced on 180nm. The fact that flipper and gekko are so simple makes them extremely good candidates for 65nm. That would put you at about 13% of the original core size. Meaning you could make almost 8 cores with the same amount of silicone. So the 'brains' of the Revolution should be significantly cheaper than the 'brains' of the Gamecube. Another serendipitous effect of the die shrink would be the ability to clock higher. It would be baffling if IBM couldn't at least tripple Gekko's core clock on a modern process. Three times Gekko's 485Mhz gives you about 1.45Ghz. 1.45Ghz on a 90nm process (let alone a 60nm one) with a 20 million transistor chip shouldn't be too challenging... Ditto on Flipper. 2 or 3 times flipper on a modern process is absurd. If you triple the clocks of Flipper you have a <i>scorching...</i> 486Mhz. Yep, 486Mhz and this on a chip that is a quarter as complicated as the simplest chip ATI is currently producing (RadeonX1300 is 105 million transistors and is clocked upto 600mhz).
So with the exception of the RAM performance is therefore easily tripled across the board just by a die shrink... The thing is Nintendo has invested a good chunk of cash in R&D. I'm expecting a die shrink and somewhat substantial enhancements. Between the two it seems Nintendo would almost have to TRY to artificially limit performance not to at least triple the specs of the Gamecube.
So with the logic above, doesn't it seem safe to believe that once the developers get their final dev kits and info on the graphics system we will have a system that, with the exception of RAM is over 3 times more powerful than the gamecube? Being a tech nut, particularly in regard to GPU's it boggled my mind to read,
Exact clock rates were not disclosed, but one development source we spoke to had this to say of the Revolution CPU and GPU: "Basically, take a GameCube, double the clock rate of the CPU and GPU and you're done."
I was boggled not because I'm a ****** who expected XBox360 performance in the space of a few DVD cases, but because process shrinks alone should do much, much more than just double the clock speed.
Thanks for reading this,
-Kevin
My mind was boiling at the idea of releasing a 'souped up XBox' today. In fact if you read all of this thread you'd have seen my post saying pretty much the same stuff I said above several pages ago. I've since made my peace and decided not to believe the IGN specs.... again not because I'm a ******, but because the progress of technology does not support their claims.