There are different kinds of swap:
disk swap and video card-system
RAM swap. Ideally, especially if you are talking about SSDs, you want as little disk swap as possible hence and focus be between video card and system RAM swap which can slow things down if the program isn't written correctly and is limited tot eh bus bridge between the two.
It was asked, about 18 months ago, "
Can the GPU use the main computer RAM (as an extension)?"
And the main answer was:
"Short answer:
No, you can't.
Longer answer: The bandwidth, and more importantly, latency between the GPU and RAM over the PCIe bus is an order of magnitude worse than between the GPU and VRAM, so if you are going to do that you might as well be number crunching on the CPU.
CPU can use a part of VRAM (part mapped into the PCI aperture, usually 256MB) directly as RAM, but
it will be slower than regular RAM because PCIe is a bottleneck. Using it for something like
swap might be feasible."
This is what I mean about chicken-fox thing (forgot about the corn). For people who don't know the premise is
A man has to get a fox, a chicken, and a sack of corn across a river. (This is the swap between RAM and VRAM)
He has a rowboat, and it can only carry him and one other thing. (PCIe)
If the fox and the chicken are left together, the fox will eat the chicken. (graphic perforce problems)
If the chicken and the corn are left together, the chicken will eat the corn. (large disk swap occurs)
Unified RAM doesn't have to do that tag team nonsense and can trade off the RAM total as needed.
