DieAppleDie said:
Exactly, why is people always leaving out of the equation those high band 32mb eSRam? |
Because it is 32MByte only? How much can you do with 32MByte? The first batch of games will use the eSRam as framebuffer and call it a day. What if the game uses 1G of textures? The better developers will start to use parts of the eSRam for data cache and variable storage purposes, but it will take time to figure out how to do this and the developers will have to program the technique for it and manage all the data.
The latency problem.. oh my, how many time has this horse beaten to death now? Do people really think that Sony engineers weren't aware of higher gddr latencies? The not so surprising answer is: Of course they were. That is why they added lots of transistors into the SoC. Some of the added features we know (direct command bus), some we don't. And fortunately for us, some engineer had a bright moment years ago and invented something called "cache" and "cache controller". The end result is roughly the following: a) for code fetches: the latency problem simply does not exist at all. b) for data fetches: Here, we might see the latency problem sometimes. How often depends on how well the programmers have defined the data structures. With some good thinking, I'd guess than in less than 1-5% of all fetches we might see a latency. it is clear, assuming the unveiled specs are correct, that the 8G gddr5 outperforms the 8G ddr3 + 32MeSRam.
The gpu's were chosen to do what they are supposed to do: deliver 1080p games (or tvtvtvtvtvsportssportssportstvtvtv for the XBox One). They contain the best parts available for the money alotted to graphics. Anybody expecting a GeForce680 in a console lives in a dreamworld.
And one last request: please stop the "the cloud will save it" messages. Not.going.to.happen.
