freedquaker said:
It's funny because I agree with everything technical you said above but obviously you misunderstood some of the points I have made, and have a different perspective than I do. I never said Esram had anything to do with GDDR5; I said, MS went with 8 GB DDR3 + ESRAM, instead of 4 GB DDR3 + 4 GB GDDR5. The first one takes A LOT MORE space on the die than the second one. It's not like ESRAM is incompatible with GDDR5 but if you'd choose to GDDR5, then it becomes pointless to use ESRAM at all. So MS had two options (given that they didn't know about the GDDR5 yields)... a) Employ 8 GB DDR3 + 32 MB ESRAM, which takes a lot of space, so you also sacrifice Compute Units and bandwidth b) Employ 4 GB DDR3 + 4 GB GDDR5, which takes much less space, so you don't sacrifice Compute Units or bandwidth - Option a does not cost less than option b, possibly costs even more! - Option a has less bandwidth, less compute units, has less power overall Why is "a" less powerful than "b" overall? Because it has ESRAM in it! You may argue otherwise but let's try different configurations c) Employ 4 GB DDR3 + 4 GB GDDR5 + 32 MB ESRAM, what's the point of having ESRAM with GDDR5 on board? Useless and only wastes precious GPU. d) Employ 4 GB DDR3 + 32 MB ESRAM, Not enough RAM etc...
Regardless of how you look at it, employing ESRAM creates a bottleneck because it is prohibitive of using a better RAM solution like GDDR5. You'd never use ESRAM with GDDR5, not because they are incompatible, but because the use of GDDR5 negates (unnecessiates) ESRAM. Either way, MS never considered GDDR5 to begin with. That's the evil beneath. |
Nah, they couldn't do DDR3+DDR5, I don't think AMD has ever made a memory controller that handles different types of DRAM in conjunction with each other, it's always been either/or pretty much.
If they wen't in your theoretical direction... Microsoft would need to retain the same high density memory chips, but because there will be half of each type, the memory bus would also have to be cut in half from 256bit to 128bit.
If they decide to retain the 256bit bus, they would then need to cut the memory density and throw twice as many memory chips onto the PCB, which will drive up PCB complexity, so costs would have increased.
Plus, keeping things simple is more ideal from a manufacturing point of view, if there was a shortage of DDR3 or GDDR5, then it would impact every console, that's twice as likely to occur with two different memory technologies.
It would have been more ideal, I agree, but cost is one of the primary factors in a cost sensitive device.
Besides... 256bit DDR3 can be faster than 128bit GDDR5 memory anyway. :)
Sony took a gamble, it paid off, Microsoft just tried to play it safe and it didn't pay off, they were more focused on the software side anyway, which also fell short on launch, for everything that was done well in the UI, there is another that brings it down.
--::{PC Gaming Master Race}::--