NVIDIA GT200 Rumours & Speculation Thread
- Thread starter Arun
- Start date
- Status
Just don't tell'em the amount of VRAM - odds are, nobody would believe you.
Possibly, it may be 4GB VRAM.... that would make nobody believe
2GB wouldn't be that surprising, considering the targeted market, and the fact that there are 1.5GB G80-based cards out there for some time now.
As a Geforce, however, a single-GPU 2GB~4GB high-end card would be somewhat of a novelty (yes, i know there are 2GB 9600 GT's, but it's not the same thing since they won't even remotely use all of that space anyway).
Given that no 2gbit gddr3 memory exists, would be indeed hard to believe - would need 32 1gbit chips. I think the pcb layout is already complex enough with 16 memory chips...Possibly, it may be 4GB VRAM.... that would make nobody believe
2GB OTOH doesn't seem outrageous - while 1gbit gddr3 chips seem to be a bit rare and only available up to 1Ghz (it seems the fast stuff is only available for 512mbit chips), several vendors have these in their portfolio (and there are some cards today using them already).
Anticipation... is everything!
The local memory is just the parallel data cache that's used for inter-thread communication. It's there in G80 as well.
You're probably not going to have luck finding any insights by comparing different interpretations of G80's structure. There probably isn't a single diagram that shows everything.
The diagram mentions there's no real representation of all the memory pools. The GT200 diagram talks about those though
And this one?
I still don't see the local memory...
The IU could be the interpolator unit?
IU is likely that, yeah.I might revisit the G8x diagrams after GT200 to add some extra detail back in.
It seems that nVidia uses some kind of invisible pixies to perform texture adressing
http://www.forum-3dcenter.org/vbulletin/showpost.php?p=6568820&postcount=523
Nothing revolutionary. Certainly nothing that should get Rys so excited. Wonder what he's talking about....
Nothing revolutionary. Certainly nothing that should get Rys so excited. Wonder what he's talking about....
It seems that nVidia uses some kind of invisible pixies to perform texture adressing
I don't know about that... If ATi's managed to squeeze 32 TMUs and 800 SPs into the ~276mm die of RV770 I'd say they're the ones with the magic
Love_In_Rio
Veteran
2 times the G92 transistors and 1.5 times G92 performance is an efficiency increment ?
Given the extra oompf apart from sheer fps - yes.
- Status
