Thoughts on 2 different PC configurations

If I was editing, I would buy a pre built PC and especially from HP. The Intel i7 4790k would be enough for video editing and should do it efficiently. The 32GB RAM will help with the multi-tasking and rendering it fairly quickly.

If you're talking pure video editing, the GTX 780 is overkill. I just had this very long conversation with another person looking for a similar PC.

http://www.tomshardware.com/forum/id-2284368/choose-video-editing.html#14097740

Along those same lines, this will probably be a good system for you.

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Xeon E3-1231 V3 3.4GHz Quad-Core Processor ($247.98 @ SuperBiiz)
CPU Cooler: NZXT Kraken X31 69.5 CFM Liquid CPU Cooler ($73.98 @ OutletPC)
Motherboard: ASRock H97 PRO4 ATX LGA1150 Motherboard ($88.98 @ SuperBiiz)
Memory: PNY XLR8 16GB (2 x 8GB) DDR3-1600 Memory ($139.00 @ Amazon)
Storage: Crucial MX100 512GB 2.5" Solid State Drive ($209.99 @ Adorama)
Video Card: MSI GeForce GTX 770 2GB TWIN FROZR Video Card ($312.98 @ Newegg)
Case: Nanoxia Deep Silence 2 ATX Mid Tower Case ($99.99 @ Micro Center)
Power Supply: SeaSonic 620W 80+ Bronze Certified Fully-Modular ATX Power Supply ($74.99 @ Newegg)
Optical Drive: LG GH24NSB0 DVD/CD Writer ($13.99 @ Newegg)
Operating System: Microsoft Windows 8.1 Pro - 64-bit (OEM) (64-bit) ($128.98 @ OutletPC)
Total: $1390.86
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2014-09-03 13:02 EDT-0400

Here is a quick quiz for you:

1. What part(s) of your workload will take place in the GPU (aka video card?).

2. When you say the difference seems significant, what measurement are you using?

Also of interest is what software you're using or proposing to use.

If the video editing software you are going to use is Sony Vegas, go with an HD 7970, R9 280x, or an R9 290. There might be others that perform similarly, but I do know that Sony Vegas and AMD GPU's do well together, from the benchmarking I have seen. I would probably go with something more like this. I included a PSU, Blu-Ray, SSD, and Win 7 pro to the list. Price was 1575 with only the SSD added.

PCPartPicker part list / Price breakdown by merchant

CPU: Intel Xeon E3-1231 V3 3.4GHz Quad-Core Processor ($247.98 @ SuperBiiz)
Motherboard: ASRock Fatal1ty H97 Killer ATX LGA1150 Motherboard ($97.99 @ Newegg)
Memory: G.Skill Sniper Series 16GB (2 x 8GB) DDR3-1866 Memory ($148.50 @ Newegg)
Memory: G.Skill Sniper Series 16GB (2 x 8GB) DDR3-1866 Memory ($148.50 @ Newegg)
Storage: Crucial M550 256GB M.2-2280 Solid State Drive ($159.98 @ SuperBiiz)
Storage: Seagate Barracuda 3TB 3.5" 7200RPM Internal Hard Drive ($99.00 @ Amazon)
Storage: Seagate Barracuda 3TB 3.5" 7200RPM Internal Hard Drive ($99.00 @ Amazon)
Storage: Seagate Barracuda 3TB 3.5" 7200RPM Internal Hard Drive ($99.00 @ Amazon)
Video Card: Sapphire Radeon R9 290 4GB Tri-X Video Card ($420.98 @ SuperBiiz)
Case: Corsair SPEC-01 RED ATX Mid Tower Case ($54.99 @ Newegg)
Power Supply: EVGA SuperNOVA NEX 650W 80+ Gold Certified Fully-Modular ATX Power Supply ($79.95 @ Amazon)
Optical Drive: LG WH16NS40 Blu-Ray/DVD/CD Writer ($54.99 @ Newegg)
Operating System: Microsoft Windows 7 Professional SP1 (OEM) (64-bit) ($141.27 @ TigerDirect)
Total: $1852.13
Prices include shipping, taxes, and discounts when available
Generated by PCPartPicker 2014-09-03 14:22 EDT-0400

The playback issues you are having, are unlikely related to the video cards render pipeline size. The GTX560Ti SIMD instruction pipeline is certainly not the bottleneck for real-time playback of your video. I guarantee this.

The likely possible sources of your interactive real-time playback issues while editing video are:
1. CPU: Not enough CPU power to muscle through the transitions in real-time. The performance of this frame-to-frame rendering of transitions and effects and manipulations are still dominantly CPU bound, even in applications with openCL and/or CUDA acceleration. The advanced parallelism of a GPU doesn't scale well until we remove the real-time limitations of a play-back and instead render/encode the whole project. In cases where you can leverage the parallelism of a GPU, the GPGPU performance of the $150 GTX750Ti is often on par with high end ($300+) Kepler based products, yet it dissipates ~1/2 to 1/3rd the power. A GTX780 just doesn't make much sense here.
Assuming your bottleneck is CPU related, it's hard to beat the value of the E3-1231V3 on a C226 chipset motherboard with ~16GB of ECC memory for this sort of work. Though changes to software settings may also help. Reducing the real-time playback quality for editing, or adjusting automatic preview rendering settings may help immensely.

2. DISK/RAM/Buffering: If you're working with high bit-rate footage the disk you are working from may not be able to sustain the data rates necessary to pull from multiple sources simultaneously through camera transitions. Remember, the data rate doubles during 2-source transitions in real-time playback. There are also many common editing mistakes that could lead to unnecessarily high disk access issues. Make sure you don't have unnecessary layers with bleed-through content (like 1% transparency instead of simply editing it out).
Assuming this is the problem, changing the way the program buffers and scratches and pre-renders may be a solution. Increasing the amount of RAM and disk space available to buffer previews may help. The best solution in a video editing environment is to use a many-disk strategy. Use an SSD for your system/software/boot drive, and then use high density (modern) mechanical drives for video editing. At bare minimum it's a good idea to have one drive configured for the storage of the original raw footage, and another for all writing/output/render/scratch operations. I would advise against having any other programs like an MP3 player making read calls on your editing drives, use a separate drive for your personal media/storage. Modern 2TB drives seem to be the sweet spot for performance/cost/reliability.


3. Fixed Function Video Decoder Performance: This is unlikely the issue, but if you are working with high bit-rate or high resolution or high FPS (or some combination of these things) footage, and if the software you are using leverages the fixed function video decoder/encoder, it *could* be a bottleneck. It's important to understand that more expensive GPU does not automatically mean better video decoder hardware. Go here: http://en.wikipedia.org/wiki/Nvidia_PureVideo ... Note, the GT520 shares the exact same fixed function video decoder as a GTX TITAN. $40, $1000. There is no difference in video playback capabilities between these 2 GPUs when using their hardware decoders (there are some limitations for post-processing on the GT520 but that's another issue that doesn't require a $1000 GPU to solve).
If the fixed function video decoder is the problem, changing software settings to avoid using it, and instead use software playback may solve the problem, however, the best solution here would be to replace the GPU with a GTX750, which is 2 generations ahead of the GTX560Ti in terms of the video decoder (VP6 vs VP4).
FYI: The fixed function video decoder in your GTX560Ti, is comparable in performance to that found on an R9 290X.
FYI: The fixed function video decoder of a GT520, GTX TITAN, are the same, VP5. Doesn't make sense to buy a GTX780 to get this feature when you can get a GTX750ti with VP6 for a lot less money and a lot less watts.
FYI: The Fixed function video decoder found on a Haswell i5/i7/E3 is stronger than any decoder/encoder hardware found on any existing discrete GPU. Some video editing software under some conditions will actually run better with a haswell integrated GPU, than with a $500+ discrete GPU. The E3 Xeons with this feature have model numbers (not to be confused with version numbers) ending in 5 and 6. (IE E3-1246V3).
Whether or not you should even be concerned with the "video" capabilities of your "video card" depends heavily on the formats you are shooting in and to what degree your editing software will even leverage it. In most cases, professional video editing doesn't even use the fixed function decoder/encoder hardware, everything is handled by the CPU/software (and openCL/GL/CUDA where applicable for effects/adjustments), which produces the highest quality results.

Adobe CC supports AMD gpu's as well. It uses Open CL instead of Cuda.

Hi jsunww,

1. I wouldn't buy into an E5 V2 at this time. Haswell E is scheduled to release soon. If you wanted to go "all out nuts" on a video editing workstation the new haswell 6-8 core chips would be the way to go.
The E5-2620V2 you mentioned is slower than an FX-8350 in video editing and costs twice as much. That wouldn't be useful.
The Ivy E parts worth consideration would be stuff like the E5-1650V2. Costs twice as much as the E3-1231V3, and offers about 25% better performance in video editing. Tough sell for me. I wouldn't bother going multi-socket for a single user application unless you can find evidence that there is useful performance scaling (probably won't scale to multiple separate CPUs at all).
The E5-1650V2 offers about 25-30% better performance than the E3-1231V3 for twice the cost. The E5-2697V2 offers about twice the performance of the E3-1231V3 in video editing at 10X the cost. As you can probably figure performance scaling and price scaling are not on the same track here. Once you cross over into E5 territory, you're paying a premium for compute density, a feature that only carries value in actual enterprise applications where every square inch of enterprise chassis is expensive and every square inch saved by improved compute density can command a huge price premium as it is quite literally only having to compete with the cost to add more machines and all of the costs associated with those additional machines. For most single user environments, the "consumer/workstation" platform offers the best value. For now, that is the 1150 socket platform, and I would advise considering the C226 chipset with ECC memory for such a build. 16GB is plenty for video editing.

2. Using a separate hard drive for the footage from each "camera" is an option, and might help, but I can not advise on whether or not changing your drive configuration will improve editing performance any further without knowing what sort of bit-rate you are recording at. Most consumer cameras are 50mbit or less, at which point, hard drive performance isn't going to be an issue unless you're working with something really ancient. Most modern mechanical drives can do ~1Gbit or better sequential read/write speeds, so should have no problem keeping up with the real-time playback of many layers of 50mbit content.

3. GPU performance scaling is hit and miss depending on the content you are working with and the specific filters/adjustments/effects applied to the content. For many users working with 1080P content, there will be no substantial difference between a GTX650 and a GTX780Ti when editing video. For folks working with 4K content, there is more useful performance scaling with beefy GPUs as it's easier to scale into the increased parallelism with larger data sets to work on. I think the 750Ti marks a sweet spot for value here. If you don't mind me asking, what benchmarks are you basing your idea to use a GTX780 on?

The 970 is the better buy, but I would only advise getting into such an expensive GPU for video editing if you have specific plans to leverage GPGPU accelerated trans-coding, otherwise, there isn't much useful performance scaling beyond a GTX650.

Definitely go with the GTX 970 over a GTX 770.

Whether or not there would be any benefit to setting up a big RAID array is dependent on the bandwidth requirements of the footage you are working with. Ideally, the drives used to hold your media (unedited raw footage) should be able to deliver at least 2-4X the bandwidth of the format you are recording in, so that they can comfortable play-back through transitions/blends etc. Modern Mechanical drives with high density platters are capable of sustaining well over 1Gbit/s data rates. Amateur HD work is ~50Mbit, so building a raid array to work on 50mbit files would be absolutely pointless, as the single drive is already capable of ~20-30X the data rate of the footage. On the other hand, if you're working with 4K footage with minimal compression (various raw-like formats), you could need anywhere from 2-10Gbit/s bandwidth just to support playback of a single video stream, which could mean that you would need a RAID configuration good for anywhere from 4-40Gbit/s to provide a good "working" speed for editing.

Point being, there's no way to answer whether or not your drive configuration idea is an intelligent solution for your needs without knowing more about the bandwidth of your footage, or what you might intend to "grow into" in the future. I will make a recommendation though: If you're planning on getting serious, look into convergent design recorders.

--------

On another note: You should put the GTX560Ti that you already own into the new build an see how it does before throwing money at a new GPU. You're making a common mistake here in assuming that the thing in the computer calls the "video" card must have a LOT to do with "video editing." The "video card" really should be called the "display generator with SIMD pipeline and video decoder." The role that a "video" card plays in video editing is not as significant as the name would imply.