I've been thinking a bit about the path Microsoft ought to take with their hardware after the fact with Natals release. It doesn't seem to make sense to release new hardware in 2011 or 2012 because taking two quick steps in relation to hardware like that is something that Sega would have done back in the day and that never worked out for them. It may never be accepted by the market as it could split the userbase where Xbox Live intercompatibility is so very important.
Then thinking about 2013 and 2014, the former is 3 years after Natal which may be both too late and too soon at the same time. Its too late to respond to either Sony or Nintendos actions. They still need to position themselves appropriately from next year for both the new technologies like 3D and improving TV technology. In addition to this they may need to increase the memory allocation in the operating system of the Xbox to truely take advantage of the new interfaces possibilities. The problem? 3 years may be too soon for the users adopting Natal to see new hardware come out.
I remembered something from a blog based in Seatle where MS-land is and that was 3 completely new hardware revisions on the table after Valhalla or even replacing that concept. I also thought about it strategically and it seemed to make sense that if they were getting AMD to go deep within the bowels of their chips to unify them, they can easily add some new features and up spec as they go at the same time. I also remembered comments on 'forward compatibility' and how Microsoft wanted to move to more itterative hardware design rather than taking big steps.
I suggest you ignore this next part if you don't like technical stuff.
A Jasper motherboard, note the spaces and try to imagine what it would look like if you delete the chip on the right which is the GPU and Ed-Ram on the one package.
As they stand with Jasper the die sizes and processes are as follows:
- CPU 135mm^2 on 65nm
- GPU 121mm^2 on 65nm
- ED-Ram 64mm^2 on 65nm
- 8x 512mbit GDDR3
To give an estimated shrink you simply take the ratio of the squares so 65^2/80^2 = 0.66 then add a little bit because not all shrinks are perfect so lets take .72 scaling in this example. In that case the ED-Ram would shrink from 64mm^2 to 46mm^2. So now we have all the chips on the same process node we can then look at what size the whole system would take up. Im going to cut it down a further 20mm^2 to account for them reworking the chip and the deletion of input/output hardware.
CPU + GPU + (shrunk) ED-RAM -20mm^2 = 302mm^2
Now performing the same operation as before 45^2/65^2 = 0.47 and like before not perfect scaling so I will assume a perfect shrink as the previous shrinks have left a large margin within the dies. ch gives us 141mm^2. To check my working the original die size total was 438mm^2 and a direct shrink from 90nm to 45nm at the same scaling rate would be theoretically 0.25 (104mm^2) and thats pretty close to the final result as well if you do not assume a perfect shrink.
You can rejoin here if you're not technical
What would they want to do with the Xbox Wii'd that they cannot do with the Xbox 360?
The Xbox 360 is 5 years old next year, so the values that the device was made for have shifted in that time. Now the machine has to meet a whole new set of values and take full advantage of a completely new interface. Full HD televisions are coming into vogue, 3D at cinema is having a revival and this will translate at some point to home users and to get full use out of an interface which works off the 3D positioning of limbs it would be best that the actual console can display 3D images as well.
Im not saying that the current Xbox 360 cannot take advantage of Natal in games, all im saying is that it cannot take full advantage of all the possibilities in Natal and waiting 3 years for new hardware seems a little foolish as that would simply give Nintendo and Sony all the time they could need to examine and beat Natal. With what was the Valhalla revision they now have an opportunity to give the Xbox 360 the hardware legs it needs for the next 4 years or so.
Doubling the ram and trippling the memory bandwidth would cost them less per unit than keeping their obsolete and outdated ram, increasing the die size of the combined CPU + GPU + ED-Ram is still much cheaper than the two package, 3 chips they use at the moment and give them room for further cost optimisation. In addition modern GPU architectures pack in more performance per watt or per mm^2 than the older 360 architecture. A little more of something newer goes a long long way in comparison. If they can render everything at full HD 1080P they can drop the scaler and give 3D images the crispness they need.
Tease.
Essay Pro