If Sega's upcoming announcement (well I figure they will eventually announce something, since they licensed it) contains at least specifications, there could be some food for pure theoretical conversations and nothing else.
And yes dksuiko is right; deferred rendering will never be able to prove itself without a real high end offering for the PC. Series5 had an initial triple target: console/PC/arcade, of which only the last has been met and I don't see anything else coming from it either.
Small IHVs just don't seem to have the resources for timely releases, irrelevant of how high my hopes were in the past. Add to that the fact that IMG has to deal with various other markets it's addressing (quite successfully I might add), then it's rather natural that such a company will dedicate it's major strength wherever the biggest income comes from.
Cross-examples being S3 and XGi; DeltaChrome was abysmally late and Volari was completely underwhelming. The latter was supposed to release a SM3.0 part last year too, but I guess they also realised by now that it'll not only come late, but that it won't be anything but another mainstream part.
With IMG I personally rather prefer either a high end part on time or nothing; we've had mainstream/budget parts before.
I've always seen TBDR as a viable alternative; as with any other architecture it obviously has advantages and disadvantages. If they've managed to solve the problematic cases - which I personally think they have - and you get on average predictable performance then I don't see there being any problems.
IMRs aren't what they used to be either with a vast oversimplyfication. Instead of running into bottlenecks where the architecture shows it's negative sides, IHVs have found workarounds for those.
An early Z optimized application is nothing else but "application driven deferred rendering", an IMR gets there via the application, while a TBDR through it's driver (not entirely correct in the strict sense, there's hardware at work in both cases).
ATI started out with quite simplistic hierarchical Z and look how it has evolved nowadays and what other techniques they've added in the meantime; same goes for NV too. What makes you think that the PVR folks have been idle in further improving their architectures too? I can see "hierarchical tiling", display list compression and what not patents and I'd guess that a healthy bulk of those are already present in MBX. Yes I know the next best argument will be that it's a small embedded chip; it doesn't seem to choke from "high" (in relative terms for the size of those chips) level of geometry; I'd rather say that the competition has to prove itself in that department right now. Ironically for the first round the only other serious competition are this far Bitboys and Falanx. BB uses an IMR and while Mali is tile based I still don't have a clue if it's a DR.
And yes dksuiko is right; deferred rendering will never be able to prove itself without a real high end offering for the PC. Series5 had an initial triple target: console/PC/arcade, of which only the last has been met and I don't see anything else coming from it either.
Small IHVs just don't seem to have the resources for timely releases, irrelevant of how high my hopes were in the past. Add to that the fact that IMG has to deal with various other markets it's addressing (quite successfully I might add), then it's rather natural that such a company will dedicate it's major strength wherever the biggest income comes from.
Cross-examples being S3 and XGi; DeltaChrome was abysmally late and Volari was completely underwhelming. The latter was supposed to release a SM3.0 part last year too, but I guess they also realised by now that it'll not only come late, but that it won't be anything but another mainstream part.
With IMG I personally rather prefer either a high end part on time or nothing; we've had mainstream/budget parts before.
I've always seen TBDR as a viable alternative; as with any other architecture it obviously has advantages and disadvantages. If they've managed to solve the problematic cases - which I personally think they have - and you get on average predictable performance then I don't see there being any problems.
IMRs aren't what they used to be either with a vast oversimplyfication. Instead of running into bottlenecks where the architecture shows it's negative sides, IHVs have found workarounds for those.
An early Z optimized application is nothing else but "application driven deferred rendering", an IMR gets there via the application, while a TBDR through it's driver (not entirely correct in the strict sense, there's hardware at work in both cases).
ATI started out with quite simplistic hierarchical Z and look how it has evolved nowadays and what other techniques they've added in the meantime; same goes for NV too. What makes you think that the PVR folks have been idle in further improving their architectures too? I can see "hierarchical tiling", display list compression and what not patents and I'd guess that a healthy bulk of those are already present in MBX. Yes I know the next best argument will be that it's a small embedded chip; it doesn't seem to choke from "high" (in relative terms for the size of those chips) level of geometry; I'd rather say that the competition has to prove itself in that department right now. Ironically for the first round the only other serious competition are this far Bitboys and Falanx. BB uses an IMR and while Mali is tile based I still don't have a clue if it's a DR.
