By using this site, you agree to our Privacy Policy and our Terms of Use. Close
HappySqurriel on 10 April 2012
Edit Thread
Squilliam said:
HappySqurriel said:
Squilliam said:

Do the math. How many transistors do you realistically expect that they put in the Wii U given a 45nm process and relatively normal die sizes? The Xbox 360 CGPU is 168mm^2. Even doubling that with better transistor scaling would likely only yield ~1.3B transistors @ 1.5* greater density on the same process.


A 45nm process would allow for 4 times the transistors on their CPU/GPU at a similar price to the XBox 360 CPU/GPU at launch being that the XBox 360 launched with a 90nm process; we also only have the CPU manufacturing process being confirmed, and (being that AMD has migrated towards the 28nm process for their high-end GPUs) it is possible that the GPU has moved to a 32nm or 28 nm process which would allow for a processor that had 8 to 12 times as many transistors relative to the XBox 360 GPU at launch ...

Beyond that, I'm talking real-world performance not theoritical performance. Simply updating the bus on the CPU and GPU and moving to 2GB of GDDR5 rather than the 512MB of GDDR3 that is currently in the XBox 360 would result in significant improvements in real world processing power of the system without substantial improvements to the processors; and this improvement can be amplified by introducing a 16MB to 64MB eDRAM on the GPU to store the screen buffer and act as a high-speed texture cache. Part of this improvement that would come from this is increased efficiency because the memory is faster, but the largest impact would be that artists could produce far more detailed environments because they were less bound by what could fit in memory; a large portion of the reason so many assets in PS3 and XBox 360 games are relatively low detail is not that the GPU can't render the image fast enough with them at a higher detail, it is because if they were higher detail they would take up too much memory when they're off-screen and result in data being reloaded from disc.

Xbox 360  volume: 6,600cm^3

Wii U volume: 1,200cm^3

Xbox 360 production cost at launch: North of $300

Wii U production cost at launch: South of $300

Cost per area of silicon: 28nm > 90nm in todays currency at their respective launches.

You can argue all you want about how great the Wii U could be. The physics don't support your argument, nor do financial constraints. If Nintendo were producing a console which cost $500 your arguments would start to make some sense as you could consider things like vapour chambers for cooling and more liberal use of chip binning.

Chances are good they are using DDR3 for cost and power reasons. Chances are good they aren't putting >300mm^2 of silicon into it. Chances are good they are setting the clock speeds at a level to maximise yields and reduce costs. Chances are good that they will employ a standard cooling solution. Chances are good they won't let the chips hit 90 degrees C as has been typical of laptop designs.

 

 


You're making a ton of assumptions, and then using those assumptions to make your conclusion ...

In theoritical floating point performance the Cell processor in the PS3 is several times as powerful as the Xenon processor in the XBox 360; and in real world in-game performance the two processors are fairly similar. While processors are ultimately bound by their theoritical performance, how well suited a particular processor is to running a particular application really determines the performance of that application on the processor; this is why DSPs with several orders of magnituded less theoritical processing power can often outperform high-end CPUs in tasks like video decoding.

While the Xenon is a good processor it is far from the be-all and end-all of gaming CPUs; even using the same manufacturing process on the same size of silicone die running at the same clock-speed, a processor could be produced that is substantially better suited to modern games than the Xenon and have substantially higher real-world performance. Although I wouldn't expect Nintendo to do this, you could create a co-processor for dedicated physics that wouldn't be substantially more expensive or power hungry than a Radeon HD 4650 and off load a substantial amount of the work off of the CPU and increase overall-performance; and, with AMD's hybrid crossover technology, it would make sense to use a low end GPU as a physics processing unit to allow it to be used to enhance graphics for games that are not heavily physics based.

Nintendo doesn't have to break the laws of physics, they simply need to design a system that is better suited to gaming than the XBox 360 to see substantial improvements in real world processing power.