i7-5960X is far better. Dual video cards in SLI is a waste of money and gives no performance gain over a single video card.
Here's the problem with your plan:
1) You selected the wrong CPU in the planned new build. That particular i7 is permanently locked to default speed (this means that it cannot be overclocked at all, and the motherboards will not let you do so even if the boards themselves have overclocking features), and its default Turbo speed of only 3.7 GHz with all four cores in use (the 4.0 GHz maximum is with only one core in use) is nowhere near high enough to compensate for the loss of four cores and eight threads. By contrast, most i7-5960x CPUs are easily overclockable to 4.2 or 4.3 GHz. Thus, all other things being equal, a quad-core CPU needs to run at 5.6 GHz in order to roughly equal the performance of an octo-core CPU at 4.0 GHz.
2) The Socket 1151 CPUs still have a limitation of 16 PCI-e lanes total from the CPU itself - and thus, going dual-GPU will result in each GPU receiving only eight lanes instead of the 16 lanes that the GPUs themselves normally occupy. This will result in GPU-accelerated performance becoming slightly slower than if both GPUs each receive all 16 lanes that they normally would have used.
3) Going SLI will not increase the total amount of VRAM - but instead, will result in exactly the same amount of available graphics memory as a single GPU. This is because in dual-GPU SLI systems, half the VRAM serves only to mirror the other half of the VRAM. In addition, SLI has been known to cause severe stability issues in Adobe productivity programs. And even if you use two GPUs in a non-SLI configuration, the VRAM on one card cannot be used for anything that's rendered on the other GPU; thus, two 4 GB cards in a non-SLI configuration will result in a 4 + 4 GB configuration instead of a true 8 GB configuration.
Now, your other plan (Xeon) is also flawed:
1) Most Xeons run at far lower clock speeds than their i7 siblings. If you choose the wrong Xeon, you'd end up with a CPU that runs at a clock speed of only 2.4 GHz or even lower.
2) Xeons are only worth it if you go for two identical Xeon CPUs and a dual-CPU-capable motherboard. A single Xeon is never worth the cost in terms of performance.
In other words, don't waste your money on an entirely new PC. Especially since a PC that performs sufficiently better than your current i7-5960x setup would have cost tens of thousands of dollars (USD) - far more expensive than what an i7-6700 based setup (or most Xeon-based setups, for that matter) would have cost. As such, there is absolutely no cost-effective upgrade at all whatsoever from your current i7-5960x rig.
So you recommented use i7-5960x and 2 video cards in non SLI mode for better performance?
No. NEVER SLI the two video cards for the Adobe applications! SLI will cause major problems there.
okay 5960x and 1 GPU :thumbsup:
Actually, I was confused about your last post. I re-read that response again, and two video cards in non-SLI mode it is. This will provide improved performance. However, you might not need two video cards for your i7-5960x system if you go for a GeForce GTX 980 Ti or a Titan X.
what about server motherboards with 2x XEON processors? that can be better then 5960x config?
Only if you get the most expensive CPUs in the Xeon line. However, if you're going to run Premiere, be aware of the 32-core/64-thread limit total. Two of the less expensive Xeon E5 series CPUs would be no improvement over your current single 5960x - and might actually be a downgrade performance-wise while costing you more money overall.
just interest why PS not use CUDA cores? i think CUDA cores give better performance for big projects in PH, Premiere, Autodesk products and etc....
This is because Photoshop does not use the CUDA API by itself (although some third-party plugins for PS do use CUDA). It uses only OpenCL, which NVidia gaming cards do a relatively poor job in.
And in Premiere (Pr), once you go over 500 CUDA cores, the graphics memory bandwidth becomes the limiting factor: On a video card with 1664 CUDA cores, a card with 256-bit GDDR5 memory running at an effective (for CUDA applications) throughput of 168 GB/second will noticeably outperform an otherwise identical card with 128-bit GDDR5 memory running at an effective 96 GB/second throughput.
And when I stated that Pr is limited to 32 cores and 64 threads, I meant CPU cores and threads, not GPU cores and threads. That means that it will use those particular maximums even if your system has more than those amounts in the CPUs.
tnx you a lot!
...you did not mention which GPU you currently are using and other details about your storage setup, memory speed, and program settings.
To achieve best performance in PPro with your current system :
1. Overclock the CPU......many users are getting up to 4.5Ghz for a large boost to performance.
2. OS, programs and Windows page file ONLY should go on your SINGLE SATA III boot drive, like a minimum 256GB Samsung 850 Pro.....an SSD RAID 0 for ALL files may not be working that great. Once loaded, programs and OS are "memory resident" and do not need the speed of a RAID 0,or, PCI SSD.
3. Place ALL other files : media, project,previews,cache,media cache, and exports on a super fast Samsung 950Pro PCI SSD for best performance. This new drive is BI-DIRECTIONAL, meaning it will read AND write at the SAME TIME, unlike SATA drives which can only do ONE of these operations at a time. In addition, the read speed is over 2GB per second and the write speed around 1.5 GB per second. That means reading is FOUR times faster that a good SSD and writing is THREE times faster. Of course, you could also RAID 0 THREE or more quality SSDs instead of the using a 950Pro, but, best is the PCI SSD.
4. Make sure that your memory allocation is properly set in the PPro settings.......the minimum recommendation is to have at least 3GB assigned to each core, which in your case would only use 24GB.....so, you would probably assign a lot more, maybe 6GB per core, I am not sure.
5. Because your CPU is so fast, if you use an NVidia 980ti or better GPU, you would want to make sure that the speed of your system memory is fast enough for the rest of your components. Eric Bowen has written about this on the forum....he can advise.
You may want to visit PPBM7.com to test your machine using PPro on the video editing benchmark test where you can compare its performance to other machines and also find out if you have any "bottlenecks". After making improvements, you can test it again to see how the performance may have increased.