What CPU and GPU hardware leverages Power Director Ultra 11's parallelism?

metazone21

Newbie Joined: Aug 19, 2012 00:15 Messages: 22 Offline

Solved by

Dec 18, 2012 09:22

1. I'm looking to buy PowerDirector Ultra 11 and am building a computer for it. I'm trying to determine what CPU and GPU I should put in. Is the parallelism on the GPU or the CPU more important? The PD ad shows a computer w/ Intel Core i5-3570 CPU (w/ integrated Intel HD Graphics 4000 – note that the ad said the Intel Core i5-3550 but that must be wrong because that has the Intel HD Graphics 2500 GPU). It has a 2nd GPU, the nVidia GeForce GTX 680, which has CUDA blocks for additional parallelism.
a. How much parallelism for video rendering is PD Ultra 11 written for – does it distribute across the 2 GPUs and then across the CUDA blocks?
b. How much parallelism for CPU processing is the s/w written for – I think most s/w is written to not take advantage of more than 4 cores on a CPU so … would going for a CPU with 6 cores (and Hyperthreading) give much of a gain for PD Ultra 11?
c. Bottom line: which of the following would result in better video editing performance:
• Configuration 1: Better GPU
=> CPU: Intel Core i7-3770K Processor (8M Cache, 3.5-3.90 GHz, 4 cores, Intel HD Graphics 4000)
=> GPU: nVidia GeForce GTX 680
• Configuration 2: Better CPU
=> CPU: Intel Core i7-3930K Processor (12M Cache, up to 3.80 GHz, 6 cores; no integrated graphics card)
=> GPU: nVidia GeForce GTX 570
d. So … would the second configuration be better, with a faster/better CPU but with no second (integrated) graphics card but using an nVidia card that has multiple CUDA blocks for video-rendering parallelism or would the first one be better with a CPU that’s probably 25% slower, only 4 cores, but uses 2 GPUs (and the second one has CUDA blocks for additional parallelism

2. I read that CyberLink PowerDirector supports hardware acceleration on the nVidia cards; is that the same as taking advantage of parallel video rendering using CUDA blocks?

3. One site recommended nVidia Quadro 4000 for the GPU for Video Editing. Is it much of a performance gain over the above GPUs?

vn800rider

Senior Contributor Location: Darwen, UK Joined: May 15, 2008 04:32 Messages: 1949 Offline

Dec 18, 2012 10:05

These may be worth reading

http://forum.cyberlink.com/forum/posts/list/25731.page#140088
http://forum.cyberlink.com/forum/posts/list/25978.page

http://forum.cyberlink.com/forum/posts/list/25134.page#136769
http://forum.cyberlink.com/forum/posts/list/25826.page#140632

Cheers
Adrian Life is really simple, but we insist on making it complicated. (see below)
Confucius
AMD Phenom IIX6 1055T, win10, 5 internal drives, 7 usb drives, struggling power supply.

BeFi

Newbie Joined: Dec 16, 2012 16:17 Messages: 17 Offline

Dec 18, 2012 15:35

Hi metazone21,

I just reverse-engineered the same question havong bought/built my system already and trying to get the most out of it.

See my findings in the following post http://forum.cyberlink.com/forum/posts/list/26138.page

My comparison seems to provdie strong evidence that the integrated GPU is much more efficient than a powerful discrete GPU.

I presume that the higher-end Intel GPUs (i.e. HD4000) are faster than their lower end siblings (i.e. HD3000) in the lower-end core i's.

With Intel QuickSync my ASUS Z77V-Pro / Intel core i3770 Win7 64bit-based system is stunningly fast in editing and rendering HD video and rock stable. I have not overclocked any of my hardware. So I can recommend this combination.

Cheers
Bernhard

metazone21

Newbie Joined: Aug 19, 2012 00:15 Messages: 22 Offline

Dec 18, 2012 23:20

@Adrian - thanks very much - I especially found the PDF (multiGPGPU.pdf) helpful ( http://forum.cyberlink.com/forum/posts/list/25826.page#140632.

@Bernhard - thanks very much as well. I looked at Jeff's analysis (tested PD 11 using GPU=GTX 580 and CPU=i7-3770 (w/ HD4000 GPU) to determine results w/ discrete GPU only, onboard GPU only, and both [multiGPGPU] on times) and it looks like it's not even worth it getting the discrete card. Note that he made no mention of needing to use Lucidlogix Virtu MVP; I'm a newbie at this so ... do you think it has anything to do with your using the AMD HD 7750 as opposed to an nVidia GPU?

Also, from your analysis, the only way to get the time down to 2.5 minutes was to have both GPUs utilized w/ PD11 added to Virtu MVP. I was wondering what your first scenario [the one that took 5 minutes)] would result in if you added the PD11 to Virtu MVP (or does that even make sense?)? Meaning, display attached to HD4000, 0% usage of HD7750, PD11 added to Virtu MVP, see PD11 switch to 'Intel Quick Sync' (does that even make sense in this scenario?) and then see what the time would be (in order to determine if a fast rendering time can be done only w/ the HD4000.

Would you mind sharing your config? I know the GPU, CPU, and MB; what RAM, case, PSU, Cooling and Storage setup (e.g. SSD? HDs in RAID?). I only take this info as input as opposed to rushing out and buying so no need to worry about giving me any advice that would result in an impulse buying decision

Again, thanks for the info --

Bill

metazone21

Newbie Joined: Aug 19, 2012 00:15 Messages: 22 Offline

Dec 18, 2012 23:27

Bernhard -- sorry, forgot to ask you which one of the ASUS Z77V-Pro MBs you have -- Which ASUS Z77V-Pro mother board do you have: Click here

BeFi

Newbie Joined: Dec 16, 2012 16:17 Messages: 17 Offline

Dec 19, 2012 02:12

Hi Bill,

let me respond to your questions one-by-one:

[Quotes] "... looked at Jeff's analysis .... and it looks like it's not even worth it getting the discrete card"
BF >>> I tend to agree. The acceleration in my fast scenarios seems to come solely from the onboard GPU. The discrete GPU only provides the preview support.

"Note that he made no mention of needing to use Lucidlogix Virtu MVP"
BF >>> Thats a big disappointment for me too. The press suggests that 3rd gen Intel CPUs provide native support for Quick Sync. It took me a while to see the first piece of SW able to use Quick Sync on my machine and only with the help of Lucidlogix.

"...do you think it has anything to do with your using the AMD HD 7750 as opposed to an nVidia GPU?"
BF >>> I can't tell.

"Also, from your analysis, the only way to get the time down to 2.5 minutes was to have both GPUs utilized w/ PD11 added to Virtu MVP. I was wondering what your first scenario [the one that took 5 minutes)] would result in if you added the PD11 to Virtu MVP (or does that even make sense?)? Meaning, display attached to HD4000, 0% usage of HD7750, PD11 added to Virtu MVP, see PD11 switch to 'Intel Quick Sync' (does that even make sense in this scenario?) and then see what the time would be (in order to determine if a fast rendering time can be done only w/ the HD4000"
BF >>> I didn't test that as PD11 always offers Hardware Acceleration even without LucidLogix. HW acceleration using the GPU to which the display is attached. I suspect that the scenario you are interested in will not give any advantages as in this case the on-board GPU HD4000 would shoulder all load. There is no headroom to be expected over the scenarios with Lucidlogix - where the onboard GPU had 80% load already.

My config:
ASUS Z77 V-PRO Deluxe mainboard
Intel Core i7 3770S
Prolimatech Genesis cooler with 2x120mm Noctua fans
16GB Memory (of which I never see beyond 4GB in use)
Seasonic x560 power supply, semi passive
Sapphire HD7750 (passive cooling)
Fractal Design R4 case, two Noctua 140mm fans front/rear
Samsung 830 SSD for Win 7 Prof 64bit
Intel M25 SSD for app data
2 WD Green 3TB each for big data
TBS Sat TV card
The result is a normally silent yet powerful system when required.

metazone21

Newbie Joined: Aug 19, 2012 00:15 Messages: 22 Offline

Dec 20, 2012 20:28

Thanks for the info - really appreciate it