Thanks for the link.
It seems like I am better off to focus on the drive IO as I am capped at about 230 MB / Sec Sustained on my drives, thanks to the lame SATA controller on the MB.
Seems like BS when they promote the board as being SATA 3 and then not be able to support it - Thanks ASUS - not.
I sense that you are not fully understanding where your current bottleneck is (weakest link) in your current build. For the workflow where you mention above when you state that you are max'ing out your CPU cores whereas your GPU is not maxed out, this is indicating that your CPU is the bottleneck - not your drives, RAM, motherboard bus, or GPU.
So, if you are not already overclocking your 3930k, you should do so. Get an excellent cooler, not just a good one, and shoot for at least 4.2GHz. When you monitor your cpu core temperature during a render you will probably not want to go much higher than 70 degrees C.
If you demand more speed still, you will need to get a more powerful cpu and/or use multiple CPUs. There are lots of Xeon cpu choices out there now with 8, 10, and even 12 cores (16, 20, 24 cores with hyperthreads considered); unfortunately, they are VERY expensive compared with the i7 models. Don't envy the 4930k, it would not help you much at all. If you wait, Intel's X99 chipset and associated CPUs will refresh the X79 options and very likely have single CPU options with greater than 6 cores.
You have already been pointed to a link (Tweakers Page) that is rich with information compiled by an experienced NLE builder that took the time to share that information in a clear concise manor - thanks Harm! You can learn a whole lot too about how your hardware performs by running the PPBM6 benchmark and comparing your composite results with other systems using similar as well as dissimilar components (see www.ppbm7.com; thanks Harm and Bill). This Adobe Premiere centric test uses different timelines to help flush out your CPU, GPU, and drive speed using separate tests.
I appreciate the detailed response.
You are right, I am still trying to determine where I am bottlenecked as it changes by task and software.
I am using an ASUS P9x79 Pro Board
I have my i7 overclocked - and it is stable at 4.6Mhz..
I have a Coolmaster Hyper 212 CPU cooler
No aftermarket GPU cooler
I have 64 GB of ram running at a modest 1600 - which I guess I could push way faster as well.
I know my I/O is bottlenecked from my testing as well.
I usually allocate 12 GB of ram to a Ram Cache with a long flush period, which has really helped me test out different I/O performance scenarios.
I've been testing a RAM drive for Cache as well, but these software approaches further task my already stretched CPU and they use RAM.
This is why I've been testing the multi-drive scenario to open up the I/O path with less CPU burden.
However, I've found my SATA 3 speeds bottlenecked at 230MB/sec despite Samsung Pro's Specs.
This is why I think the ROC from Areca or LSI is the next step - to bypass the drive controller on the MB.
However, I also see my GPU max out when running complex projects - multi-cam or when using several effects, so I am guessing a new GPU at some point as well.
But you are correct that I do seem to be CPU bound most of the time.
I was hoping to off load taks to the GPU and The ROC to unburden my CPU.
I can't afford to buy a new MB and dual Xeon system yet.
I've been spending on my Camera rigs too.
There is only so much cash to go around.
I've been assuming that a major contributing factor is Adobe's spotty use of CUDA. I find I get much better performance on certain tasks from more fully cuda optimized software. Unfortunately, these other tools are not Adobe Premiere, so I always come back - so far at least.
I'm thinking the next step is the ROC controller. I've been waiting for the PCI 3.0 options to actually start shipping before buying one.
Same for the GPU - have been waiting for the Titan Black to ship.
As well I'm eyballing a RAID array to go along with the ROC but undecided on 10Gbit Network connection that would improve my whole network or box just for my video workstation. I know the 10Gbit network station maxes out at 1100MB/Sec where as the local RAID can push 2300Mbits per second.
Lots of tradeoffs to consider.
All this is likely to cost another $5k or more.
I can't do it all at once, so I'm trying to come up with a logical implementation sequence that delivers the most bag for buck at each step.
I'm a bit green on optimizing specifically for Premiere, so I am hoping you experienced guys can point me down a road with fewer twists and turns.
I like the twisties on my VFR800, but not on my computer.
Which of these ports do you have your SSD's connected?
"Intel® X79 chipset :
2 x SATA 6Gb/s port(s), white
4 x SATA 3Gb/s port(s), blue Support Raid 0, 1, 5, 10
Marvell® PCIe 9128 controller :
2 x SATA 6Gb/s port(s), white
ASMedia® ASM1061 controller :
2 x Power eSATA 6Gb/s port(s), green"
If you were to use one of the two Intel white connectors you should get full performance from the SSD, if you use the Marvell ports or ASMedia ports I am not sure as I do not own that board. Maybe someone with experience on an Asus P9X79 Pro board could chime in here.
I am using the correct Sata ports.
I also upgraded the MB bios to latest version to make sure
I even got new cables to make sure and I'm still stuck at 230Mb/sec on sustained read and writes.
I've tried 3 different SSD drives - Samsumg Pro and Evo as well.
I've tripple checked everything and even tried all the Sata ports, just to make sure.
Maybe the board is defective?
I've asked Asus for help but I get ignored.
So I am thinking my only options short of swapping out the board is to get a RAID card.
I am open to suggestions on this one, as it has me stumped.
The only thing I have not tried is a different disk speed tester.
But I am using Crystal Disk like everyone else.