Nvidia provides the manual, and if you look at the paper manual that
came with the product, you may notice the "Nvidia font" used in the
front section of the manual. I downloaded the EVGA copy of the manual
again to have a look.

http://www.evga.com/products/pdf/680i%20manual.pdf

The DIMM slots are colored in the manual. The two blue would be dual
channel. The two black would be dual channel.

The slots are numbered like this. The manual says to use slot 0 and
slot 1, as positions for a pair of DIMMs. Or, you can use slot 2 and
slot 3, as positions for a pair of DIMMs. Those would be the dual
cnannel choices. Slot 0 and slot 1 are the same color as each other.
Slot 2 and slot 3 are the same color as each other.

     ---------   slot 0
     ---------   slot 2

     ---------   slot 1
     ---------   slot 3

One difference from using dual channel, would be that more of the
potential bandwidth of the PCI Express lanes on the video cards,
might be usable. The target for PCI Express DMA transfers, might
be system memory, so the higher the system memory bandwidth, the
more PCI Express packets can be handled per second. The theory being,
that this somehow would cause the bus arbiters to offer the PCI bus
more opportunities to do any necessary data transfers.

But really, on how many chipsets, is arbitration policy or options,
known ? The answer is not many. Sometimes, information is released
about how the bus arbitration amongst multiple buses works, when
there is a problem with the chipset.

But having just looked at a handful of threads, from a search on
X-Fi and 680i, there are lots of theories, some from previous
motherboard experiences. My own experience with sound problems,
is you can try a solution, think you have it solved, only to
find tomorrow, when you operate the computer, the snap, crackle
and pop are back. What that means is, some of the suggested solutions
provide a small amount of relief, without changing the nature of the
problem. All it takes, is some change in the background conditions
in the machine (say you use an application you weren't using yesterday),
and the problem can return.

By all means, try the dual channel setup - that is good for the
machine anyway, and will give you better benchmarks. But it may
turn out to be just a bandaid, and not really fix the sound problem.

One thing I don't understand in all this, is the practical details
of how sound buffers are passed to the sound card. It appears that
the sound card is demand based. There is a FIFO buffer of some sort.
When the FIFO drains to a certain depth, an IRQ is sent to the
processor. The remaining depth of the FIFO, and the drain rate set
by the sampling rate on the DACs, determines the maximum time before
the interrupt must be serviced. If the interrupt is not serviced, the
last sample stored in the FIFO, is reused over and over again,
creating a "flat spot" on the analog waveform. If that happens
intermittently, and many times a second, that is a "crackle".

Increasing the size of the FIFO buffer would sound like a natural
solution, but that also has consequences for lag between when sound
samples are computed or prepared, and when they are played back.
A game would be less responsive acoustically, if a whole second of
fixed sound samples had to drain out of the card, before the game
could cook up a new sound effect. So the latency caused by the buffer,
cannot be made too large, AFAIK. I don't know if there is an article
on a web site, that puts numbers to all these assumptions, but it does
seem to be a tough problem to solve. Certainly enough years have
passed, for people to have had a chance to fix it.

Using a large FIFO, would help with streamed sounds. In that case,
the values being sent to the sound card, are known well in advance,
and if you had a really big RAM buffer on the sound card, you could
put a whole song into the sound card. In that case, there would be
no opportunity for snap, crackle, pop. But for dynamic content,
such as sound effects in an FPS, a deep FIFO would be a liability.

Good luck,
    Paul