Going to get one of these when they come out

[MiRai]

edit: you didn't comment on the Gskill RAM though! 2800 Mhz!

Alright, well it's probably going to cost an arm and a leg and be completely unnecessary.

To run at 2800MHz you're first going to need a motherboard that can handle that and then you're going to have to pump a lot of voltage into the memory controller for stability. Looking at the larger screenshot it's most likely going to require a whopping 1.65v to reach those speeds which can damage the memory controller (the link is for Ivy, but also holds true for Sandy).

As for the timings, 11-13-13-35 aren't phenomenal, but the raw speed (MHz) is going to make it fast regardless. So, let's do some quick math as per this post on the Anandtech forums:

Let's just do the math, the frequency is expressed in Hertz, which means "cycles per second". So, the DDR3 2133 will perform 2133 cycles a second while the DDR3 1600 will do, well, 1600. You, of course, know this.

Now the CAS latency is given in cycles. So, a CAS8 DIMM will take 8 cycles to respond and the CAS11, 11 cycles.

Now putting it all together - the DDR3 2133 CAS11 will take 11/2133 seconds, which is equal to 0.00516 seconds, to respond while the DDR3 1600 CAS 8 will take 8/1600, which is equal to 0,005 seconds, to respond. Thus, the 1600 DIMM is faster. For your dilemma, you're contemplating 0.000516 versus 0.00422 or timing difference of 0.00094 seconds! How fast are your reflexes and how long are you willing to wait for your memory to respond? Yes, I'm being an ass. You'll never ever see, feel or sense a difference.

You'll also want to carefully check the timings, since the higher the MHz the looser they are. I've seen 2133 DIMMS with 11-14-28-30, simply junk. As well as voltage, avoid anything higher than 1.5V. The only way many manufacturers get a respectable CAS at high MHz is to crank the voltage to 1.65V. Hence is why the 1.35V Samsung DIMMS are so smashing. Also, less DIMMs generally means less load to the memory controller. Therefore, 2x8GB is better than 4x4GB.

11/2800 = 0.00392 seconds

Looking through some 32GB RAM kits on Newegg there isn't anything as fast (and that was expected because nothing comes close to 2800MHz), but while staying within the voltage standard of 1.5v, you can find these which are 7-8-8-24 @ 1600MHz (and have an access time of 0.00437 seconds (7/1600)) or these which are 9-9-9-24 @ 1866MHz (and have an access time of 0.00482 seconds (9/1866)). Both of these kits I've listed will work on 90% of motherboards on the market today without any funny business, while that new G.Skill kit may not.

So, the 2800MHz kit will be 0.00045 seconds faster than the 1600MHz RAM which safely stays within Intel's specifications. Are you going to notice that difference outside of synthetic benchmarks? Nope. Are you willing to shorten the lifespan of your CPU by pumping 1.65v into the memory controller for a 450 millionth of a second increase in speed?

I'll close with a quote from a DDR3 memory scaling article on Anandtech:

The results weren't very stimulating, were they? Just as expected, gaming with faster memory just doesn't make any notable difference. I could have potentially lowered the resolution and settings in an attempt to produce some sort of difference, but I felt that testing these games at the settings they're most likely to be played at was far more enlightening. If you want better gaming performance, the GPU is the best component to upgrade—no news there.

EDIT: To add, here's a list on Newegg of the 32GB memory kits running at 2400MHz (all 1.65v). Looking at those prices, I would assume that G.SKill's new kit would cost close to $400 (if not more).