Re:Yes (Score 4, Informative)

You are confused with some of your figures. Each SPU has access to 256K of local cache. This is very very fast memory... no misses. They can't directly access main memory, so there isn't really contention. However, the programming model is more complicated. Nothing console devs can't handle ;) The 16MB/s figure (if that's correct, don't remember exactly...) is related to reading VRAM (sony confusingly calls this local memory) from the Cell, something you don't generally have any reason to do. The GPU has fast access (~20GB/s) to the Cell's memory (called Main Memory) which is much more useful. You can split up your texture usage between the two memory pools and cut down on memory contention. Anyhow, too many winter brews, so I'll stop rambling.