Digital Foundry has a comparison with a convincing argument - Sony made a gamble that paid off. MS committed to 8GBs early on which seemed cost prohibitive, so had to go with local storage again. Sony chose GDDR5 and was looking at 4 GBs, but was able to stretch to 8GBs for long-term advantage. Had MS picked GDDR5 and expected 8 GBs to be cost effective, it'd be a different story.
We really need to know what the advantages of ESRAM are. It seems very wasteful at the moment to go with that over eDRAM. A local store to make up the BW deficit of DDR3 is one thing, but committing 2 billion transistors to the task instead of 300,000 seems a major decision. Of course there's the foundry issue to factor in. We can't be sure which is the greater influence - better performance or better economies of production.
We really need to know what the advantages of ESRAM are. It seems very wasteful at the moment to go with that over eDRAM. A local store to make up the BW deficit of DDR3 is one thing, but committing 2 billion transistors to the task instead of 300,000 seems a major decision. Of course there's the foundry issue to factor in. We can't be sure which is the greater influence - better performance or better economies of production.