Sorry, I meant DX11 API rather than DX11 GPU. I can't see how they'd port their tile-based deferred shading algorithm to a system without DirectCompute/shared memory, so they'd have to revert to traditional deferred algorithms which would be very bad for performance unless they reduce the number of lights (which might have gameplay implications) - I'd tend to believe it's more likely they're using CS4.x but I'm not sure. I'd love to know more details about which DX11 API-only features they're using.
Oops, correct
It's easy to lose track of how long it took for OpenGL to get on the same level as DX11! It's a pretty good API nowadays though obviously.Agreed. I suppose the main short/mid-term approach will be porting existing AAA games but even that is very difficult due to the lower performance and especially touchscreen controls.
I can't remember which developer mentioned this (might have been repi from DICE on Twitter actually), but I agree that the next big thing on handhelds has got to be high precision eye tracking. If you can actually get it to work reliably (high resolution image to analyse and proper compensation for the eye naturally scanning around all the time etc.) then that's a completely new control scheme which could be an order of magnitude better than touchscreen for games (while still allowing touchscreen controls at the same time where it makes sense). But then there's also a question of where best to process that data; ideally handheld GPGPU should be mature enough by that point to do it on the GPU but we'll see
But now I'm getting even more off-topic so I'll just stop here. Then again this isn't really that off-topic in a sense; while I was talking about SGX, the exact same points apply for Tegra 4: the lack of ES3 compliance isn't a deal breaker and the increase in performance remains the most important element for this generation of hardware. If you give me a choice between ES3 and framebuffer compression on a bandwidth starved IMR, I'll always choose the framebuffer compression.
