I have a simple problem.
I wrote some pretty straightforward GL code for walkthrought.
On my GeForce 2, it seems that enabling textures slows down a lot the drawing (as expected).
My configuration is
32 bit rendering
32 bit texture
mipmap linear-mipmap-linear
(kind of the worst case for everything).
When texture are drawn (< 16b MB, so fit in memory and bounded with glBindTexture) the system slows down by a factor of 2.
The questions are
when using non mipmaps, the slowdown is pretty much the same - why?
if in windows control panel, i say to use 16-bit texture, is it actually doing it
what is the best 16 bit texture format
what anyone can suggest to speed upo my little application
I have a simple problem.
I wrote some pretty straightforward GL code for walkthrought.
On my GeForce 2, it seems that enabling textures slows down a lot the drawing (as expected).
My configuration is
32 bit rendering
32 bit texture
mipmap linear-mipmap-linear
(kind of the worst case for everything).
When texture are drawn (< 16b MB, so fit in memory and bounded with glBindTexture) the system slows down by a factor of 2.
The questions are
when using non mipmaps, the slowdown is pretty much the same - why?
[/b]
mipmaps takes more memory amd deciding wich one to use quite simple, so there won’t be much of difference. Just watch your memory, if you have 10 256X256X32 textures map and you mipmap then you’ll have lots of joy!!
[b]
if in windows control panel, i say to use 16-bit texture, is it actually doing it
what is the best 16 bit texture format
what anyone can suggest to speed upo my little application
Thanks a lot, and sorry for the lengthy message.
–x[/b]
For the 32bits to 16bits, if they say so it should do it ! For the best 16 bit texture format, I don’t know.Probably(4 bits for each color)
As for speeding up your application, if the framerate is really bad like you could do frustum culling, you could sort your surfaces by texture.
It also depends on how many polygons you are displaying at once. If you are texture mapping a object that has say 20K polygons compared to a 10k polygon object, then you can see why it would slow down.
There is only so much bandwidth available, and as said above, it also depends on mimaps, and other factors.
Originally posted by zangel: NVIDIA says that if you are using NV_register_combiners extension then enabling mipmaping decrease texture performance at half.