Hi Pinky,
Regarding 90 nm...
Well my reasoning behind their support of 65 nm in 2005 is the following:
PlayStation 3 would not be launched before mid-late 2005 and it has been stated by Sony that their goal with PlayStation 3 is to really push the barrier, to exhibit an even better jump than the PSX to PlayStation 2 technological jump.
If you design a chip and you want to start manufacturing with a certain process, let's say 90 nm, you might still decide to push it and take some loss due to the relatively "massive" die size and you have to cope with the other characteristics that the particular manufacturing process impose: frequency will also be limited.
You might later on move the chip to 65 nm and save on manufacturing costs, but you cannot improove its performance, add execution units, e-DRAM or increase the clock frequency... the console, the PlayStation 3, has already shipped to customers and you cannot change the chip.
If you designed the chip and had 65 nm in mind as you know 65 nm will be available by the time you start mass-production then you can follow the physical limitations, you can use the transistors' budget of a leading edge 65 nm process and you will not have to take away all those execution unit, all of that e-DRAM and you can keep the frequency a bit higher... also you might need not to push as much 65 nm as you would have pushed 90 nm...
Basically you will still loose some money on the 65 nm chip as its size won't be optimal, but you will have a much better chip and it won't be as big and tough to manufacture... then you can move to 45 nm or less when it becomes available and svae on costs...
My point is that for a CPU as ambitious as the one Sony might expect to fit in PlayStation 3 with 90 nm we would butcher its specs and performance: making the chip work, stripping it down from ideal architecture to 90 nm will cut its features and performance a bit too much...
Of course the chip would have still to be "stripped down" from its ideal configuration to fit the 65 nm manufacturing process specifications, but we might be able to fit more of the essential stuff...
Do you see a nice amount of e-DRAM on a 90 nm chip ? I do not...
Also there are some other issues...
As you know Sony, IBM and Toshiba did revise their plans from 100 nm to comprehend technologies up to 50 nm and Sony and Toshiba have worked hard on sub 70 nm manufacturing technologies and in December of last year they completed the research on their new 65 nm manufacturing process and by mid 2004 they would have been able to start manufacturing 65 nm chips ( or so a pretty happy Toshiba executive was announcing back then ).
I can provide quotes for those two last pieces of information if you want, but I am sure you have read the relevant Press-Releases already.
Sony has also moved PlayStation 2 core manufacturing off Japan ( mainly to China IIRC ) and I do not think they will leave their fabs idle... this could be a good chance to upgrade them to new technology... again 65 nm needs to find implemetation, research on it has been completed and succesfully.
Another interesting thing can be found in a recently posted thread on B3D: it talks about a new fab Toshiba wants to build and Sony's involvement...
Sony, Toshiba to tie up in new chip plant - Nikkei
Fri April 18, 2003 11:05 PM ET
TOKYO, April 19 (Reuters) - Consumer electronics giant Sony Corp 6758.T and Toshiba Corp 6502.T , Japan's largest chipmaker, will join hands in developing a new generation of semiconductors in a bid to take the lead in the field, Japanese media reported on Saturday.
Financial daily Nihon Keizai Shimbun (Nikkei) said the Sony group, the country's biggest user of chips, would invest abut 50 billion yen ($417.8 million) in a state-of-the-art semiconductor plant planned by Toshiba.
The two electronics giants would jointly build the facility to mass-produce a new microprocessing unit -- the key component of digital home electronics -- at Toshiba's Oita factory, and planned to have it in operation by 2004, the paper said.
Toshiba and the Sony group decided to tie up from the construction stage of the new facility in order to begin mass production at an early date, the Nikkei report said.
The new plant would mostly manufacture a new microprocessing unit now being developed jointly by Toshiba, Sony's videogame unit, Sony Computer Entertainment Inc (SCE) and IBM Corp IBM.N .
The new chip would be used in SCE's next-generation PlayStation game console, as well as high-performance home computers and other products.
Something does not tell me they will upgrade them to 90 nm when they also said they expect by 2004 as well to start 65 nm chips production...
Intel is taking its sweet time with Prescott's launch and it is not going to kill current Pentium 4 sales and will gladly wait to kill the next Athlon's launch ( well Opteron and Athlon 64 )... however they are expected to launch Prescott in 2003, as you say Q3 probably is the Quarter that will definately see Prescott...
That is mid-late 2003 then for Intel's 90 nm, high-marging ( something tells me they are not going to sell it below cost
), mas-manufactured Processor.
Forward to mid-late 2005 ( say late 2005 )... 2 years after and we expect PlayStation 3's CPU to debut using 90 nm ?
Well either that would disastroulsy butcher its performance or it would butcher it a bit less, but it would cost a lot to manufacture and we would not even enjoy ultra good yelds...
They will have 65 nm working by mid 2004 ( the Toshiba designers expected the date to be March 2004 ), why should not they target PlayStation 3 CPU for 65 nm ?
Yes, the Emotion Engine and the Graphics Syntehsizer did launch using 250 nm, but it is also true that they had bad problems with GS manufacturing and that the cost of both processor was enormously high considered the massive size of those two chips and Sony and Toshiba were not extremely happy about it either... Sony and Toshiba have both learned a few lessons from that experience.
I believe that current problems for some with 90 nm have convinced them to research more 65 nm and to jump to that technology.
Still major players like IBM and Intel have not had THAT many problems to delay 90 nm to 2005...
Toshiba has also been working with 90 nm for SRAMs and more ( gotta find more links on this though )...
