GameIndustry.biz wrote up a fairly extensive opinion piece about the power of the Wii U, have a read of this excerpt:
In theory, we could look at laptops here as an example of getting powerful chips working in smaller areas. The problem here is that high-power mobile GPUs are highly ‘binned’ – they’re the pick of the production crop of processors destined for a broad range of different graphics cards. Mobile parts are typically the very best chips, the cream of the crop, capable of great performance at low voltages. Nintendo would not have this luxury on a mass-production item with a single design, where high efficiency is the key to keeping costs down as much as possible.
Realistically, short of a major architectural shift to components based on smartphone tech – and lots of it – the idea of Wii U possessing next-gen rendering capabilities doesn’t make a lot of sense. We know that there’s no transition to mobile tech because the IBM CPU is an off-shoot of an existing line and the firm doesn’t make mobile CPUs. Similarly, while AMD has produced smartphone GPUs, none of them get close to the performance of the Xbox 360′s Xenos GPU. That being the case, the chances are that it’s a customised variant of an existing PC Radeon part: Japanese sources have previously hinted at a connection to the Radeon HD 4000 series – and a lower-end chip from that range would be a good fit.
With the IBM chip confirmed at a 45nm process – the same as the current Xbox 360 – the question then moves on to how the graphics chip is made. TSMC, the most probable candidate for actually producing the chip, has just moved onto a 28nm process, and will be ramping up production throughout the year. But any new node typically starts with low production yields, so Nintendo would need to either swallow the cost (Microsoft did this at the launch of the 360 with the then state-of-the-art 90nm Xenos GPU) or downclock the chip. It’s far more likely that sticking to the existing, established 40nm process for AMD GPUs would actually be cheaper for them in the short term – and would provide cost-savings in the future when the chip could be shrunk economically.
But let’s assume that Nintendo does push the boat out here. Even a 45nm CPU and a 28nm GPU in a box that small is still likely to cause cooling issues for an actual “next-gen” 360 beater. The more probable 45nm CPU/40nm GPU combo combined with the size of the machine suggests a far more likely scenario: that Wii U has a ballpark performance level with current PS3 and Xbox 360 titles, perhaps actually lower. Across the years, chip designs may have become more refined and efficient but it’s worthwhile to point out that almost all major increases in processing power have mostly come from shrinks in the fabrication process meaning that more transistors can be packed into the same amount of silicon.
The final nail in the coffin about a notional 2x increase in power over the Xbox 360 comes from Nintendo itself. At no point has the platform holder ever suggested that Wii U offers that kind of leap in processing power, an extraordinary omission considering the amount of money Nintendo would need to invest in this architecture. The focus of the platform holder’s message is of course on where the money has been spent: the tablet controller, with its zero latency link to the console – technology that must have been fairly expensive.
But is there anything in the package that could give the Wii U an advantage over the PS3 and Xbox 360, aside from the tablet controller? We should look at the commodities that have collapsed in price over the past few years, and could prove genuinely useful for a games machine. RAM is the obvious choice: a 1GB minimum wouldn’t break the bank and would help developers significantly. The pre-E3 rumour of 8GB of flash RAM also makes sense, especially when we bear in mind that there is no internal hard drive. The Wii U optical drive – almost certainly based on Blu-ray technology – could also be faster than its PS3 equivalent too. This may be useful bearing in mind that the lack of HDD would preclude mandatory installs.