ninjablade said:
|
You don't know that. The 22.4GB/s on the 360 is theoretical, and it is not able to get to this peak. In addition, it uses half for reads and half for writes, so theoretical max of 11.2GB/s for reads, the WiiU can use the full 12.8 for reads (if this is the actual number, developers have suggested real-world performance appears to be much higher). Since most of the writing to the RAM can be done when a scene or level is loaded, the RAM usage would be very read heavy during properly coded games (and would be greatly optimized by the customized memory controller), meaning this 'bottleneck' may not be the huge issue you're making it out to be. You also aren't taking into account latency, which is a huge factor. Nintendo has said the memory system is very low latency, and developers have confirmed this in their praise of the memory architecture. The cache memory on the GPU and CPU's is also much larger and faster than that on the 360, meaning cache misses resulting in lost real-world FLOPS are greatly reduced or eliminated (and the OOOE on the CPU further reduces cache miss performance loss for CPU tasks, a major issue holding back the performance on the 360's CPU compared to its theoretical performance), another reason why a straight-line comparison is a moot point. Also, there is a more efficient, modern memory controller onboard (meaning the theoretical max is much closer to real-world, and that R/W optomizations are used to maximize bandwidth and cache usage, eliminating or greatly reducing the theoretical bottleneck you describe), as well as hardware texture compression that also goes against your point. Future 'ground-up' games for the system will prove me right.
