I posted the same question before:
http://forum.cyberlink.com/forum/posts/list/43132.page

In that post I asked the same stuff you did (would it be worth the cost)?

I know that not everyone is using the same source files and produces the same output format - and how much difference the card will make depends a lot on that and the rest of your hardware, but I could also not find any benchmarks with PD13 that would tell me if it would be worth it.
My processor(s) are very good, so I am not convinced that spending $200+ will make any real difference.

This is something that should come from Cyberlink.
Recommended systems - not minimum hardware requirements only.

AlS - I have 12 cores (2 cpu's) and it uses all of them during produce (actual cores - 24 with hyper threading).

https://cpubenchmark.net/cpu.php?cpu=Intel+Xeon+X5670+%40+2.93GHz&id=1307&cpuCount=2

Regarding you question on more memory. Open your task manager while you edit or produce videos. If you don't reach above 80% of memory usage then I think it would not make any difference to upgrade your memory. My computer has a lot of memory - PD uses only a small amount (I also don't do any 4k). I use big chucks of my extra ram as ram drive Win8.1 Pro x64 / Dual x5670 / 24GB / GTX960 4GB / 240GB SSD + 640GB HDD / PD13 Ultimate

I remember reading that the intel quick sync is better than a nVidia $400 gpu card in that it was just seconds behind nVidia in encoding performance and of course, it's free. That was 3 years ago.

Look at this link where intel quick symc is the best in this test: https://obsproject.com/forum/threads/obs-benchmarking-1080p-60fps-cpu-vs-nvenc-vs-quick-sync.15963/ . It depends on what you want to do. I like to use smart rendering with any video editing software. It seem like most people here prefer hardware accelleration as most can't tell the difference in their video.

nVidia and CUDA support have recurring issues in PD13. nVidia has focused on optimizing their CUDA rather than using the aptly named OpenCL standard which is, well, open. Although OpenCL does work on nVidia cards (mostly, and not always in PD13), the performance is in many cases vastly inferior to AMD GPUs.

The best bang for your buck, strictly in PD13 (the A-word's premier piece of software has gone whole-hog for nVidia's proprietary implentation) is, as it stands, an AMD GPU. Something from the Hawaii series, preferably the R9 290 (not the X, the price premium is not commensurate with the performance increase) does wonders in PD13.

The high end of the R9 series is superior to all but the higher end nVidia Quadro cards in terms of pure OpenCL performance, and can be found for under $300USD (in some cases under $250 depending on where you live). nVidia has focused on gaming, and gaming only for their consumer card architecture, hence the greater suitabilty of AMD cards for everything from Bitcoin mining to video encoding. If you spring for two and disable Crossfire in the AMD drivers, PD13 will even utilize both cards for both rendering and editing, basically giving you over 5,000 compute cores worth of theoretical possible performance, for less than the price of a single GTX 980 with a decent non-reference cooler. To sum that up, if you aren't going to spend a few thousand on a pro-grade nVidia card, don't bother with an nVidia card for OpenCL performance.

As to CPU cores, PD13 will use basically whatever is available. You will probably see an increase in performance moving to an i7, even with an identical physical core count due to the efficiency of hyperthreading - the i7 may be a quad core, but hyperthreading effectively brings it closer to an 8-core in terms of performance in most workloads. My older i7 laptop (dual core hyperthreaded) does amazing work compared to my similar i5 dual core laptop. The increase however will most likely not be as significant as you saw in the move to an SSD (unless you spring for a major CPU upgrade, say to a 5820k or 5960x).

As for RAM - more is almost always better, assuming you are loading enough content to fill it - moving from 16gb to 32gb if all you are working on is 2 minute Youtube spots probably won't give you all that much benefit - if you are working on 20+ minute 1080 or 4k content, you pretty much can't ever have too much RAM. That being said, 8gb is pretty much the bare minimum. At a guess, your system is probably using up at least 3-4gb just in system processes. 16gb should be considered the baseline for working with full HD video.

Anyway, that's my two cents.

Good luck with whatever your PC choice may be - any questions, feel free to ask!

Edit, yes I am using R9 290s myself - I came from using a pair of nVidia GTX 680s, then to a 980 which died an early, watercooling related death. I can honestly say that even using only a single 290, AMD's VCE codec and rendering performance far outpace the OpenCL performance on the 980, and give better quality results with almost equal speed to CUDA optimized software. Nanoxia Deep Silence 1 case | Ultra X4 850W PSU | Asrock Fatal1ty Killer 990FX | AMD FX8320@4.7gHz | Coolermaster Nepton 240m | 2x HIS R9 290 GPUs | 24gb G.Skill Ripjaws 1600mHz DDR3 | OCZ ARC 100 480gb SSD | Various WD Black and Seagate drives to the tune of ~20tb | I have been building systems for over 16 years, however my opinions can be and sometimes are incorrect. Please, call me out if I say something asinine.

Thanks guys - this is the most informative info I have seen.

EKSVid
Sorry I missed your thread. Interesting that everyone recommended you get an NVIDIA GTX 960'
I agree Cyberlink should give us recommended and minimum configs. Like HD vs 4k. Too many users on this forum complaining about PDR13 have been advised to get a new PC.
If PDR13 can use 12 cores then more is definately better!
I have the software to create a RAM drive but not sure how to use it. If I save my PDR13 project to a folder on a RAM drive will it make a big difference? What about shadow files?

tomasc
Wow that is the first benchmark I have seen for Intel Quick Sync. Thanks. It makes the question of adding discrete GPU to my Intel system even more confusing.

Salamand3r
Interesting reasons for choosing AMD over NVIDIA GPUs. That explains compatibility issues with PDR13. Also the info we get on Adobe Premier hardware benchmarks may be misleading for PDR13 due to Adobes use of CUDA vs Open GL. Finally we have some valuable (albeit costly) advice- that your AMD R9 390 "rendering performance far outpace the OpenCL performance on the Nvidia GTX 980, and give better quality results" - and cheaper. I also didn't know that two GPUs would work.
I built my PC last year and Haswell was the latest from Intel. It's now obsolete and limits my CPU upgrade to socket 1150 so I cannot use your recommended i7 CPUs which really irritates me!
I will upgrade RAM next to 16Gb - Thanks

This all points to one problem. Instead of having to learn from the expensive and time wasting "school of hard knocks" it's time for Cyberlink to come to the party. They know how their software works - and how hardware-dependent PDR13 performance is. If PRD13 gives us 4k, I must assume that their recommended "minimum system" will give satisfactory 4k results and that is just not true. Same goes for the new multi-cam. If we don't start with the correct hardware we will never be happy with their software.

I look forward to more user comments - Thanks again! Power Director 13&14 Ultimate, Photo Director 6, Audio Dir, Pwr2Go 10
Win 10 64, Intel MB DH87MC, Intel i5-4670 CPU @ 3.40GHz, 16Gb DDR3 1600, 128Gb SSD, 2x1Tb WDBlue 7200rpmSATA6, Intel 4600 GPU, Gigabyte G1 GTX960 4GB, LG BluRay Writer

Also remember power consumption of the card when looking at upgrades!
This site gives you the power connector needed for most popular cards:
http://www.realhardtechx.com/index_archivos/Page362.htm

For my computer the PSU (650w) is not easily replaceable and I have only one 6 pin connector left.
AMD cards in general take a lot more power than the nVidia cards.
So I could run a nVidia GTX960-970 (some only need one 6 pin power connector), but the AMD's Radeon R9 270 is the best I could find that needs only one 6 pin power connector from the AMD side.

Win8.1 Pro x64 / Dual x5670 / 24GB / GTX960 4GB / 240GB SSD + 640GB HDD / PD13 Ultimate

Your 1150 is NOT obsolete, I think I may have not really thought out my wording. I meant to say that you won't see as significant of an increase between an i5 and an i7 as what you saw moving to an SSD - the heavy storage reliance of video editing makes the perceptible feel much more biased towards storage upgrades.

Also, although you can't go for the massive amounts of additional cores that the 2011-v3 socket will allow, the new Skylake (not released quite yet) CPUs are also LGA1150, and while they don't look like they will see huge performance gains over Haswell, they do gain some efficiency. And you will see a difference moving to an i7, as I said, just not night-and-day like changing your storage media.

Just really, really don't get an nVidia card for rendering performance, even in software (like Premier) that supports it - nVidia's CUDA codecs can be really horrendous in terms of quality. Using OpenCL on nVidia cards yields better quality, but is not nearly as speedy or as well optimized as AMD GPUs.

As to dual GPUs working - I wouldn't bet on it in 100% of cases. As it was, I had to change some driver versions and finally stumble upon actually DISABLING multi-GPU rendering (Crossfire) in the AMD control panel before ANY acceleration would function without crashing PD13. However, having done so, both GPUs load up quite nicely and quite evenly when utilizing them in PD.

In regards to using a RAM drive - that turns a portion of your RAM into basically a hard drive (at least as far as Windows is concerned) capable of operating at hundreds of times the speed of an SSD. The drawbacks however are many - first, you lose access to the RAM in your RAM drive. So if you were to create an 8gb RAM drive out of your 16gb of RAM, you will be back to only have 8gb available for software to use. Second, you are very limited on capacity - is the 8gb RAM drive actually going to be enough for all the footage you are using for a project? Third, RAM drives are volatile storage, meaning they disappear when you power down. Now, most RAM drive utilities will automatically save the data to hard disk and restore it when shutting down and rebooting, but you are really walking a thin data loss line when it comes to data safety. Its completely up to you. I don't use a RAM drive except for random thought experiments and testing, but YMMV.

Finally - while I do agree that Cyberlink should at least list tiers of various levels as to what specs you need for what content, some of this is really self explanatory if you do more research. If one plans on using a computer for ones work or hobby, it behooves one to know at least keep on top of PC hardware. I recommend something like the Linustechtips youtube channel for quick and easy to digest tech knowledge that will stick in your brain and keep you informed of the general specs you will need for whatever PC you happen to require.

Cheers! Nanoxia Deep Silence 1 case | Ultra X4 850W PSU | Asrock Fatal1ty Killer 990FX | AMD FX8320@4.7gHz | Coolermaster Nepton 240m | 2x HIS R9 290 GPUs | 24gb G.Skill Ripjaws 1600mHz DDR3 | OCZ ARC 100 480gb SSD | Various WD Black and Seagate drives to the tune of ~20tb | I have been building systems for over 16 years, however my opinions can be and sometimes are incorrect. Please, call me out if I say something asinine.

Quote: Also remember power consumption of the card when looking at upgrades!
This site gives you the power connector needed for most popular cards:
http://www.realhardtechx.com/index_archivos/Page362.htm

For my computer the PSU (650w) is not easily replaceable and I have only one 6 pin connector left.
AMD cards in general take a lot more power than the nVidia cards.
So I could run a nVidia GTX960-970 (some only need one 6 pin power connector), but the AMD's Radeon R9 270 is the best I could find that needs only one 6 pin power connector from the AMD side.

Since the majority of PSUs now are single rail designs, in that scenario it would be acceptable to use a dual-molex to PCIe 8-pin or 6-pin connector to get the additional connector. It wouldn't be ideal, but it would work. Keep in mind that the secondary connector, much like the "+4" pins on on motherboard 4+4 power are mainly there as a purely ancilliary safeguard. Many GPUs will function with a single 6- or 8-pin connector populated. As long as your PSU actually delivers 650w (and assuming that it's an OEM unit based on how difficult it is to replace) it should (although there is an equal chance that it doesn't even come close, HP is particularly bad at using competent PSUs) be absolutely fine unless you are running something like an FX 9590 for a CPU.



Edit: even in your situation, the 270x should yield better and more reliable hardware acceleration in PD13 than the 960 for sure, and probably close to the 970. Nanoxia Deep Silence 1 case | Ultra X4 850W PSU | Asrock Fatal1ty Killer 990FX | AMD FX8320@4.7gHz | Coolermaster Nepton 240m | 2x HIS R9 290 GPUs | 24gb G.Skill Ripjaws 1600mHz DDR3 | OCZ ARC 100 480gb SSD | Various WD Black and Seagate drives to the tune of ~20tb | I have been building systems for over 16 years, however my opinions can be and sometimes are incorrect. Please, call me out if I say something asinine.

Quote: Using OpenCL on nVidia cards yields better quality, but is not nearly as speedy or as well optimized as AMD GPUs.

Post some hard encode time data for your AMD speedster, say 20 of the boats.wmv in the timeline produced to a very common H.264 1920x1080/60i 24Mbps profile, that way the rest of us can compare to this speed.

Thanks
Jeff

Quote:

Quote: Using OpenCL on nVidia cards yields better quality, but is not nearly as speedy or as well optimized as AMD GPUs.

Post some hard encode time data for your AMD speedster, say 20 of the boats.wmv in the timeline produced to a very common H.264 1920x1080/60i 24Mbps profile, that way the rest of us can compare to this speed.

Thanks
Jeff

As I have deleted the PD sample content, is there a link anyone can post? Otherwise, I am more than happy to reinstall PD13.



Edit: nvm, extracting from the installer archives.



Edit 2: Had to reinstall anyway, couldn't find the darn thing anywhere in the archive.



20x Boats.wmv, H.264 1920x1080/60i 24Mbps profile, total length 4:20:56, total encode time 1:15, single GPU. Multi-GPU crapped out on me again, unsurprisingly. I figured it would take me another hour to get it working again, and taking an hour to post results on a short encode would look...fishy to say the least. Nanoxia Deep Silence 1 case | Ultra X4 850W PSU | Asrock Fatal1ty Killer 990FX | AMD FX8320@4.7gHz | Coolermaster Nepton 240m | 2x HIS R9 290 GPUs | 24gb G.Skill Ripjaws 1600mHz DDR3 | OCZ ARC 100 480gb SSD | Various WD Black and Seagate drives to the tune of ~20tb | I have been building systems for over 16 years, however my opinions can be and sometimes are incorrect. Please, call me out if I say something asinine.

I should note that I am only speaking to personal experience on older nVidia GPUs - my stint with a 980 was too short to get a lot of data on it's OpenCL performance. That being said, if you do some quick research, there is plenty of objective benchmarking data coming to the same conclusion (even the superior DX12 API overhead performance of 2+ year old AMD GPUs vs. nVidia's GTX 980 can be used, albeit not entirely accurately, to infer low-level GPGPU performance), and even more subjective data on the quality of OpenCL vs. CUDA output. Keep in mind also that even the quite aged Radeon 7970 is faster than the GTX 980 in video encoding in many circumstances.

AMD's superior compute performance at this point isn't even up for debate - the pre-ASIC Bitcoin mining craze alone is hard proof of that. That being said, it would actually be interesting to start a thread just for benchmarking various hardware. I am actually a bit disappointed that Cyberlink doesn't have a built in benchmark function - with a leading encoding engine like theirs, it would probably be widely adopted.

Granted, the Maxwell core nVidia GPUs offer far, far better compute performance than any previous nVidia consumer-level offering, but they are not the top dogs in the majority of compute-centered benchmarks, and price-to-performance for compute, the comparison is ludicrous. They aren't even close.

Let me go on to say that I am most assuredly not an AMD GPU fanboy - these are the only team red GPUs that I have owned since the ATi 9800 Pro. I hate AMD's drivers, I hate their bloatware, I even hate their "Gaming Evolved" slogan. I built this PC just a few months back for two purposes - to be able to game at 4k, and to be able to comfortably work with video. On the gaming front, I lost some performance going with AMD. I only made the choice I did based on a couple of weeks of reading and research, and having found that nVidia's compute performance was still just barely up to par. Nanoxia Deep Silence 1 case | Ultra X4 850W PSU | Asrock Fatal1ty Killer 990FX | AMD FX8320@4.7gHz | Coolermaster Nepton 240m | 2x HIS R9 290 GPUs | 24gb G.Skill Ripjaws 1600mHz DDR3 | OCZ ARC 100 480gb SSD | Various WD Black and Seagate drives to the tune of ~20tb | I have been building systems for over 16 years, however my opinions can be and sometimes are incorrect. Please, call me out if I say something asinine.

To my knowledge, PD13 only uses OpenCL for accelerated effects, if the user selects option. For most user timelines very little of the overall content duration contains these "effects" so OpenCL performance for the most part offers the typical user very little for Nvidia. For all other regions of the timeline the basic Nvidia NVENC dedicated H.264 hardware video encoding is used.

I'm no fan of one particular GPU or another, owned both, just like working in facts as well as possible to make appropriate decisions. A standard test suite and results would be very beneficial for those attempting to spend dollars wisely. As in this case, if one’s not interested in HA encoding, no need for anything of the GTX970 level. However, it does take a very good CPU to beat it if H.264 typical profiles are of interest.

Here is my GTX970, encode time 0:48, 1.56 times faster than the R9 290.

Jeff

Quote: To my knowledge, PD13 only uses OpenCL for accelerated effects, if the user selects option. For most user timelines very little of the overall content duration contains these "effects" so OpenCL performance for the most part offers the typical user very little for Nvidia. For all other regions of the timeline the basic Nvidia NVENC dedicated H.264 hardware video encoding is used.

I'm no fan of one particular GPU or another, owned both, just like working in facts as well as possible to make appropriate decisions. A standard test suite and results would be very beneficial for those attempting to spend dollars wisely. As in this case, if one’s not interested in HA encoding, no need for anything of the GTX970 level. However, it does take a very good CPU to beat it if H.264 typical profiles are of interest.

Here is my GTX970, encode time 0:48, 1.56 times faster than the R9 290.

Jeff

Very impressive result on the 970, far in excess of what I expected based on other benchmarks out there. I completely agree that a standardized test would be great - to my knowledge there are no benchmarks out there that rely on PD. Vegas or Premier are the only widely used ones, and in Vegas the Maxwell GPUs fare quite poorly (which I based part of my conclusion on). Great to have another stat to add to my repertoire.

However, based on that, I am now questioning whether or not I am seeing more flakery, most evidence indicates that the 290 should still be performing better. Off to tweak things more! Nanoxia Deep Silence 1 case | Ultra X4 850W PSU | Asrock Fatal1ty Killer 990FX | AMD FX8320@4.7gHz | Coolermaster Nepton 240m | 2x HIS R9 290 GPUs | 24gb G.Skill Ripjaws 1600mHz DDR3 | OCZ ARC 100 480gb SSD | Various WD Black and Seagate drives to the tune of ~20tb | I have been building systems for over 16 years, however my opinions can be and sometimes are incorrect. Please, call me out if I say something asinine.

My result without any fancy GPU as a reference




Note.

With hyper threading truned on it was about 16 seconds slower.

Now, lets see the result of a GTX 960 Win8.1 Pro x64 / Dual x5670 / 24GB / GTX960 4GB / 240GB SSD + 640GB HDD / PD13 Ultimate

Would be interesting to know What percentage of GPU utilization when rendering video in PD13. I have older ATI HD7870. When I render the boats.wmv, it took about 2:24, but I also noticed with GPUZ that the GPU utilization was only 15-18% average. Maybe once it jumped to 30%. Why isn't the number higher? Would video rendering be faster if the GPU utilization was at 80% or more?

OH yeah, the Intel HD4600 is much faster (at least rendering MP4) than my ATI card and have considered removing it. Intel i7 4770k, 16GB, GTX1060 3GB, Two 240GB SSD, 4TB HD, Sony HDR-TD20V 3D camcorder, Sony SLT-A65VK for still images, Windows 10 Pro, 64bit
Gary Russell -- TN USA

Quote: My result without any fancy GPU as a reference




Note.

With hyper threading truned on it was about 16 seconds slower.

Now, lets see the result of a GTX 960

For what it's worth, you tested the wrong profile. Try again with 1920x1080/60i 24Mbps and your times will be 10-15% slower than posted is all.

Jeff

Ok, my bad

Ran it again, only 7 seconds slower. 1:47 now. Win8.1 Pro x64 / Dual x5670 / 24GB / GTX960 4GB / 240GB SSD + 640GB HDD / PD13 Ultimate

I posted the project 20x Boats.wmv.If PD, does not find the file, give the path "C: \ Program Files \ CyberLink \ PowerDirector13 \ SampleClips \ NTSC"

I expected a better time for my R9 270, (3.15 HE) inferios (3:03 CPU).

My motherboard is PCI-Express 2, ATI R9 270's PCI-Express 3.

That's where this difference or not? AMD-FX 8350 / 8GB DDR3
SSD SUV400S37240G / 2-HD WD 1TB
AMD Radeon R9 270 / AOC M2470SWD
Windows 7-64 / PD16 Ultimate

Quote: Your 1150 is NOT obsolete, I think I may have not really thought out my wording. I meant to say that you won't see as significant of an increase between an i5 and an i7 as what you saw moving to an SSD - the heavy storage reliance of video editing makes the perceptible feel much more biased towards storage upgrades.

Also, although you can't go for the massive amounts of additional cores that the 2011-v3 socket will allow, the new Skylake (not released quite yet) CPUs are also LGA1150, and while they don't look like they will see huge performance gains over Haswell, they do gain some efficiency. And you will see a difference moving to an i7, as I said, just not night-and-day like changing your storage media.

Just really, really don't get an nVidia card for rendering performance, even in software (like Premier) that supports it - nVidia's CUDA codecs can be really horrendous in terms of quality. Using OpenCL on nVidia cards yields better quality, but is not nearly as speedy or as well optimized as AMD GPUs.


If one plans on using a computer for ones work or hobby, it behooves one to know at least keep on top of PC hardware. I recommend something like the Linustechtips youtube channel for quick and easy to digest tech knowledge that will stick in your brain and keep you informed of the general specs you will need for whatever PC you happen to require.

Cheers!

Thanks I appreciate the tips including RAM drive. I can't keep up with the hardware world so I appreciate your advice. My Haswell scoket 1150 confusion comes from trying to find my way around the Intel website. In terms of my upgrade path it is currently limited to the Haswell range.

http://processormatch.intel.com/Processors/CompatibleProcessors?componentName=dh87mc

The newer i7s (6&8 core) are Haswell-E which is not 1150. But they include more powerful GPU (All the above are limited to the 4600 GPU - but other i7s go up to the 6100 gpu)

Skylake looks like it will be socket 1151 - then there is Cannonlake maybe next year.

As far as I can tell the Broadwell series due this year are also socket 1150.

Socket 1150 upgrades still appear to be limited but I may be wrong so would appreciate any other advice you might have. Power Director 13&14 Ultimate, Photo Director 6, Audio Dir, Pwr2Go 10
Win 10 64, Intel MB DH87MC, Intel i5-4670 CPU @ 3.40GHz, 16Gb DDR3 1600, 128Gb SSD, 2x1Tb WDBlue 7200rpmSATA6, Intel 4600 GPU, Gigabyte G1 GTX960 4GB, LG BluRay Writer

Quote: OH yeah, the Intel HD4600 is much faster (at least rendering MP4) than my ATI card and have considered removing it.

Thanks would be interesting to see what impact the newer Intel 5000 & 6000 GPU's might have on PDR13.

My 20 x Boats H.264 1920x1080/60i 24Mbps with my little i5 CPU and Intel 4600 graphics took 1:54

Tried again with Intel Quick Sync Vid off and got - 3:51 Power Director 13&14 Ultimate, Photo Director 6, Audio Dir, Pwr2Go 10
Win 10 64, Intel MB DH87MC, Intel i5-4670 CPU @ 3.40GHz, 16Gb DDR3 1600, 128Gb SSD, 2x1Tb WDBlue 7200rpmSATA6, Intel 4600 GPU, Gigabyte G1 GTX960 4GB, LG BluRay Writer