XBox 360 hypotheticals thread

I agree. While 30MB of eDram would be optimal for the given resolution, it was very unlikely. However, even the jump to 15MB would have been huge. No one has been able to clearly explain to me how to perform a single Z only pass @ 720p using 4xAA within the current 10MB constraint. 15MB allows for that and then subsequent passes need only 2 tiles. It also reduces pressure for multiple render targets, etc.

Hindsight being 20/20....
If 15MB of eDram was at all possible they should have gone for it. Plus beef up the texture caches and added support for fp16 blending. On the CPU, L2 cache should have been 1.5MB and run at full-speed. Then finally modified the DVD-ROM firmware to support VMD via a dashboard update. :)
 
Lack of tiling would have benefited elsewhere, such as the features that don't work with it. I can only presume though that the cost was prohibitive, because it seems a straightforward choice. Either that or they were hoping for 4xAA to be common in which case tiling would still be an issue, so they thought to save some costs and go with a minimum for adding. Alstrong's theory about BC with AA sounds good, but BC was always an afterthought. I'd be surprised if they chose their next-gen eDRAM with BC in mind when none of the rest of the system had any regard for BC!
 
Actually, Mintmaster brought up a more obvious point (that I missed :oops: ) concerning the number of users who have yet to jump to HD or even connect to an HDTV. Now I'm thinking that was the main reason, and the part about backwards compatibility was just a fortunate coincidence. :p
 
Yes, I agree that it was really chosen for 4xAA @ 480p, since it was by far the most common TV resolution during the design time frame and tiling for HD resolutions probably seemed like a acceptable compromise. It's too bad they somewhat underestimated the importance of HD. That extra 4-5MB would have made a big difference.
 
Back
Top