The key bottleneck issue is usually the memory channel and the stack/register algorithm used to minimize bottlenecks. Case in point is the Tegra 3, since they built a fast engine (chipset) into a car with bald tires (memory channel).
Any bottleneck along the process chain is same result- a bottleneck. The new Sammy chip apparently has to juggle between two quad cores, so might have weakness in some tasks.
The more I mess with the iPad 4 and its chipset, the more impressed I am, so something that is even 50% faster- real world, NOT lame (IMO) synthetic benchmarks- I will be VERY content
If even faster, butter on the biscuit!