1) True
2) Things like that are handled through drivers and hardware, not the software. But yes, thats how it should work.
3) No. Its commented out in DX for example. If you try to add it, it will end with many MAVs.
OpenGL has no such flagging (the program has to manage that) and on DX side, its better in most cases to let DX manage where the texture is.
4) No it will not be smooth as 8meg cards most likely, as those cards have a far too low bandwidth to send all those data up again and again. As well, those systems will most likely not have 100mb free ram for your textures as well and when they need to go to conventional ram, you could load them from disk nearly as fast as well.
For such systems, creating lower res texture versions (by hijacking BMs own mipmapping code) on startup would definitely be a good idea.
2) Things like that are handled through drivers and hardware, not the software. But yes, thats how it should work.
3) No. Its commented out in DX for example. If you try to add it, it will end with many MAVs.
OpenGL has no such flagging (the program has to manage that) and on DX side, its better in most cases to let DX manage where the texture is.
4) No it will not be smooth as 8meg cards most likely, as those cards have a far too low bandwidth to send all those data up again and again. As well, those systems will most likely not have 100mb free ram for your textures as well and when they need to go to conventional ram, you could load them from disk nearly as fast as well.
For such systems, creating lower res texture versions (by hijacking BMs own mipmapping code) on startup would definitely be a good idea.