No it won't, single card and SLi will run fine even on PCI-E 2.0, its when you go Tri SLI or Quad SLI that PCI-E 3.0 is needed

Honestly if you want to buy the Sabertooth you may as well wait until the Maximus VI Formula comes out (supposedly fairly soon, similar price point, and no sooner than when you'll actually be able to get 2 780 Classys) because it will allow you to buy a third GTX 780 in the future (you say you will have this rig for a while). But no, you won't bottleneck them whatsoever with PCIE 3.0 x8. That is plenty of bandwidth for current-gen and next gen cards. Also, another option is that you could buy X79 and and the upcoming 4930K, or wait until they come out and buy a 3930K/3960X on sale. With X79, you have 40 lanes of PCIE 3.0, so you really won't get bottlenecked in any way (even in 4 Way SLI). 2 cards at native x16 x16 3.0 will be the absolute best performance you can get, although x8 x8 is fine for this gen of cards. On some of the higher end Z87s, you get a PLX chip (as you probably know) enabling more PCIE lanes to allow for more cards in SLI/Crossfire. With a PLX, you may be able to run 2 cards in x16 x16 3.0, but they won't be directly connected to the CPU and they will have a 3rd party controller in between. It will increase bandwidth, but also increase latency as it is another thing for the data to pass through. For 2 of any CURRENT GEN cards, I would recommend any setup that doesn't involve 3rd party controllers. Considering 2 cards CAN'T yet FULLY POPULATE the bandwidth of PCIE 3.0 x8 x8, the extra bandwidth provided by a PLX chip is unnecessary and will slow you down (albeit minutely) due to extra latency. I believe the ASUS ROG boards with PLX chips automatically bypass them in one and two card configurations, because they know it provides optimal performance.  I honestly don't recommend keeping a computer for as long as you've mentioned, because even if you get an upcoming 4960X and 4 GTX TITANs you will not have much worth talking about in 5 years. Now, if I WERE IN YOUR SHOES, I would hold out one more year and a half until Haswell-E and X99, as that will be a refresh of Intel's X chipset (only happens every few years) and you will be able to get DDR4. If you waited until then, you would be able to have the best of the best (besides GPUs eventually) in your computer whereas if you upgraded now you may feel a bit behind that far in the future. I just wanted to help provide some options, and I hope you found this (somewhat) easy to follow and helpful!!

Solar Trans

Honestly if you want to buy the Sabertooth you may as well wait until the Maximus VI Formula comes out (supposedly fairly soon, similar price point, and no sooner than when you'll actually be able to get 2 780 Classys) because it will allow you to buy a third GTX 780 in the future (you say you will have this rig for a while). But no, you won't bottleneck them whatsoever with PCIE 3.0 x8. That is plenty of bandwidth for current-gen and next gen cards. Also, another option is that you could buy X79 and and the upcoming 4930K, or wait until they come out and buy a 3930K/3960X on sale. With X79, you have 40 lanes of PCIE 3.0, so you really won't get bottlenecked in any way (even in 4 Way SLI). 2 cards at native x16 x16 3.0 will be the absolute best performance you can get, although x8 x8 is fine for this gen of cards. On some of the higher end Z87s, you get a PLX chip (as you probably know) enabling more PCIE lanes to allow for more cards in SLI/Crossfire. With a PLX, you may be able to run 2 cards in x16 x16 3.0, but they won't be directly connected to the CPU and they will have a 3rd party controller in between. It will increase bandwidth, but also increase latency as it is another thing for the data to pass through. For 2 of any CURRENT GEN cards, I would recommend any setup that doesn't involve 3rd party controllers. Considering 2 cards CAN'T yet FULLY POPULATE the bandwidth of PCIE 3.0 x8 x8, the extra bandwidth provided by a PLX chip is unnecessary and will slow you down (albeit minutely) due to extra latency. I believe the ASUS ROG boards with PLX chips automatically bypass them in one and two card configurations, because they know it provides optimal performance.  I honestly don't recommend keeping a computer for as long as you've mentioned, because even if you get an upcoming 4960X and 4 GTX TITANs you will not have much worth talking about in 5 years. Now, if I WERE IN YOUR SHOES, I would hold out one more year and a half until Haswell-E and X99, as that will be a refresh of Intel's X chipset (only happens every few years) and you will be able to get DDR4. If you waited until then, you would be able to have the best of the best (besides GPUs eventually) in your computer whereas if you upgraded now you may feel a bit behind that far in the future. I just wanted to help provide some options, and I hope you found this (somewhat) easy to follow and helpful!!

 
 
I have read about the Maximus VI Formula, and I do like it, but I want to stick with the Sabertooth for the 5 year warranty. I only build my computers every 4-5 years, and my last computer has held up for about 4 years, with just this year having to drop the setting in some games down from high to medium, and my SATA pots starting to fail, and with that 5 year warranty it would be really nice. I would love to build more frequently but do to budgetary constraints I can not. I will be able to save up for a major build, but after that I will be low on funds for a while. I have also debated about the x79 platform and get the new Ivy Bridge-E, and that can still be an option but would $300-$500 increase in a budget be worth it for what I am doing? I could wait until Haswell-E, but I am not sure if my computer will hold up, and I am not sure my funds will last as well.
 
Thank you for your advice, it has been really helpful.

Solar Trans

If you are building for 5 years, I would definitely do an IB-E.

 
Would the IB-E processor be overkill for what I am mainly doing, which is gaming? Also, would spending that extra $400-$500 be worth it?

It may be now, but think about it: people are still using their 1366 extremes from like 2009/2010. (Almost) no one is using their consumer level counterpart, as they can't compete with today's hardware very well. IB-E should have a similar lifespan, althogh the chip set will be a couple years old. I'd recommend a 4930K and an X79 DARK+Extended Warranty. Well IIRX the only CPUs with unlocked multipliers back then were extremes, so the others became quickly obsolete but I think you get my point.

Jthorpe

_Nite_ 

Jthorpe 

Solar Trans 

If you are building for 5 years, I would definitely do an IB-E. 

 

Would the IB-E processor be overkill for what I am mainly doing, which is gaming? Also, would spending that extra $400-$500 be worth it? 

 

It will be overkill to start with, but if you plan to keep the computer for 5 years you want the best you can get. 

 

Solar Trans

It may be now, but think about it: people are still using their 1366 extremes from like 2009/2010. (Almost) no one is using their consumer level counterpart, as they can't compete with today's hardware very well. IB-E should have a similar lifespan, althogh the chip set will be a couple years old. I'd recommend a 4930K and an X79 DARK+Extended Warranty. Well IIRX the only CPUs with unlocked multipliers back then were extremes, so the others became quickly obsolete but I think you get my point.

Ok, so If I have the funds, I should go with the X79 and IB-E, but if I fall short on funds, would I still be ok with the Z78 and SLI x8/x8?

  You will be plenty fine. If you get a board with a PLX chip then it would essentially be 16x 16x. One such board is the one I am using now. Or the EVGA Z87 Classified will be capable of that when it is released. Not sure what other boards have it, but I'm sure Asus and MSI have a board with a PLX chip as well.

First products to use PCI Express 1.0a appeared in 2004 - GeFore 6800 Ultra (16 pixel pipes at 400MHz)
First products to use PCI Express 2.0 appeared in fall 2007 - GeForce 8800 GT (112-CUDA cores at 1500MHz)
First products to use PCI Express 3.0 appeared in spring 2012 - GTX680 (1536-CUDA cores at 1006MHz)
PCI Express 4.0 isn't scheduled to appear until around first half of 2015.  
[source]
 
See a trend?  The number of CUDA cores each specification seems to support has been going up by an order of magnitude each time.  The first 4.0 flagships could have over 10,000 CUDA cores in a single-GPU and probably support GPUs with 100,000 CUDA cores by the time PCIe 5.0 shows up a decade from now-- yet 4.0 cards will be backward compatible with 3.0 slots, just like 3.0 cards are backwards compatible with 2.0 slots.  
 
By the time a card comes that you would have to worry about your dual 3.0 x8, your CPU may severely bottleneck.  In other words, your old-by-then Haswell processor will betray your future graphics upgrades after 780 Classified long before your platform choice does, but your pair of 780 Classifieds are fine in your planned configuration.  
 
Just remember, time doesn't stop, your cards barely use the space provided in the PCIe 3.0 spec which is designed for growth not current needs.

Solar Trans

For now, 2 cards can't fully populate x8 x8 3.0, so PLX chips just add latency. In the future, it could bring more gain. But yes, if no IB-E (which I highly recommend) then Haswell will be OK. I would say go for the Z87 Classified on release and buy an extended warranty over the Sabertooth (no PLX or 3/4 SLI hurts it's long term value, IMO not worth the $250 even if it looks cool)

I was looking at the Z87 Classified (now that I have found out about the EVGA extended warranty , and it looks sweet), but it is an E-ATX, and the case that I would like to use (Factal Design Arc Midi R2) will not support it. I can look a other cases though, if it is getting great reviews. Also, dosen't it have a PLX chip, so increased latency?

The reason I recommend it with the PLX is because while it may hurt you minutely now, if you get more/new GPUs during the lifespan of your build the extra bandwidth will likely be much appreciated and the latency will become obsolete. ROG boards with the PLX bypass it with 2 cards, which improves FPS a little tiny bit now but in the future when you're needing that extra bandwidth you'll wish it didn't. I have a Corsair 800D and it's amazing, albeit expensive. Really, only people with multiple WC loops need the 900D, and even then the 800D is usually acceptable. All in all, if you can afford it, I recommend it!! (Although the extra money may be better put elsewhere)

Solar Trans

The reason I recommend it with the PLX is because while it may hurt you minutely now, if you get more/new GPUs during the lifespan of your build the extra bandwidth will likely be much appreciated and the latency will become obsolete. ROG boards with the PLX bypass it with 2 cards, which improves FPS a little tiny bit now but in the future when you're needing that extra bandwidth you'll wish it didn't. I have a Corsair 800D and it's amazing, albeit expensive. Really, only people with multiple WC loops need the 900D, and even then the 800D is usually acceptable. All in all, if you can afford it, I recommend it!! (Although the extra money may be better put elsewhere)

Ok, I see what your saying, but if I do not plan on getting a third GPU, or getting new GPUs in the future, then I could stick with a non-PLX x8/x8 board right? I have seen that case, and it is very nice, the only thing is that it might take too much of my budget.

I’m assuming this reviewer is looking at PCIe 3.0 x 16 vs PCIe 2.0 x 16. As I recall PCI 3.0 x 8 ~ PCIe 2.0 x 16 and so there may be some useful data here to consider.
http://www.anandtech.com/...aling-pcie-2-vs-pcie-3

To add onto a few other posts, remember one fact: With each PCIe revision, the specification calls for "doubling" the bandwidth.  Therefore:
 
PCIe x4 @ 3.0 = PCIe x8 @ 2.0
PCIe x8 @ 3.0 = PCIe x16 @ 2.0
PCIe x16 @ 3.0 = unearthly bandwidth!
 
With that, it has been shown that running a GTX 680 on a PCIe x8 @ 2.0 can be bottle necked at high resolutions (tri-monitor), and therefore you should plan on running all cards at PCIe x16 @ 2.0 speeds.  If your motherboard drops down to x8, like in 3-way SLI, you will want PCIe 3.0 enabled on your platform.
 
Think about this question: What exactly is all of that bandwidth needed for?  
 
The answer comes from knowing how NVIDIA SLI works.  SLI works by a "each card will render 1 full frame for the entire resolution."  If running at 1920x1080, with 3-way SLI, that means each card will render 1 full frame at 1920x1080 and output it over its outputs.  The same holds true even with huge 5760x1080 and 6000x1080 setups (or more with tri-27" monitors!).  Yep, each card is responsible for rendering 1 full freakin image - this is why you need > 2 GB memory on huge resolutions.  Even with a single monitor, GPU2 has to send the entire image back over PCIe to be outputted by GPU1 (where the single monitor is connected to).
 
Now, why do you need the bandwidth?  Because with multiple monitors connected to different cards, you system uses the PCIe bus to move that data from the 1 card that rendered the full 5760x1080 image, split up and sent to the other card(s) for them to output the picture.
 
So, sending over the PCIe bus is roughly the size of:
1920x1080 = ~2 million pixel image
2560x1200 = ~3.8 million pixel image
5760x1080 = ~6.2 million pixel image
 
The data gets huge, really really fast, as the resolution goes up; and, it must be sent over the PCIe bus to be outputted by the other GPUs with monitors connected to them.  NVIDIA has been mum on exactly how they do this.  But with tri-monitors and 3-way SLI, with a monitor connected to each GPU, one can guess that GPU1 will render the image, and then that image is broken up into 3 pieces.  Once third is sent to GPU2 to render, and the other third is sent to GPU3 to render.  Some open questions are: is this done simutanously/in parallel?  Or is this done synchronously?
 
With my personal experience, 3-way is just not smooth enough for fast-action games.  Too much jerkiness.  Two-way, while lower FPS, is just buttery smooth.  Part of the issue is my board cannot do PCIe 3.0 with 3-way (known for ASRocks now).