So i'm just a little confused about this. If you are running for example a 768 gtx 460, it has 768 mb of ram. Ok, but when you add another 768 in SLI it doesn't exactly mean you have 1536mb of video memory? BUT if each card is doing less of the rendering, like half of the screen per each card, doesn't that actually give the system MORE memory to work with? I only ask because some people say you can't just combine video cards and gain more frame buffer. I hope this made sense, just wanted a little insight on that. Thanks
Video card memory confusion
- Thread starter Tortuga
- Start date
dugdiamond
New member
each card has it's own memory - they are not added together and shared between the two of them.
the gpu's memory is used by it's architecture (opengl, directx, etc), for holding images and shader code.
as games and other gpu apps get better, they are starting to use more hi-res images for our eye-candy. these images are usually in the form of n^2 resources (256x256, 512x512 ... 2048x2048), and can quickly eat up memory.
a shader can only hold 8 pointers to external data (images, integers, floats, etc) at any one time, and are passed through registers similar to 'assembler'.
a card with more memory will need less intensive data swapping with some clever coding
)
the gpu's memory is used by it's architecture (opengl, directx, etc), for holding images and shader code.
as games and other gpu apps get better, they are starting to use more hi-res images for our eye-candy. these images are usually in the form of n^2 resources (256x256, 512x512 ... 2048x2048), and can quickly eat up memory.
a shader can only hold 8 pointers to external data (images, integers, floats, etc) at any one time, and are passed through registers similar to 'assembler'.
a card with more memory will need less intensive data swapping with some clever coding
each card has it's own memory - they are not added together and shared between the two of them.
the gpu's memory is used by it's architecture (opengl, directx, etc), for holding images and shader code.
as games and other gpu apps get better, they are starting to use more hi-res images for our eye-candy. these images are usually in the form of n^2 resources (256x256, 512x512 ... 2048x2048), and can quickly eat up memory.
a shader can only hold 8 pointers to external data (images, integers, floats, etc) at any one time, and are passed through registers similar to 'assembler'.
a card with more memory will need less intensive data swapping with some clever coding)
Lol, you might need to dumb it down for me. Its not combined shared memory BUT each card does use its OWN memory to full potential? At the end of the day what i'm asking I guess, is that you DO in fact have more video memory with a second card? Even if its not combined, you still have another card in there splitting the rendering on screen....right?
dugdiamond
New member
YES
each card will have almost half of the workload - and hence, NOT share inputted resources.
both cards work on rendering part of a 'scene', with the primary card having to 'also' put the two parts together (post-rendering)
each card will have almost half of the workload - and hence, NOT share inputted resources.
both cards work on rendering part of a 'scene', with the primary card having to 'also' put the two parts together (post-rendering)
