|
|
Quote
Quote
i contacted customer service to see if i could get in touch with an engineer to resolve this issue, but it seems they think this is an nvidia problem....
Typical pass the buck. THis is definitely a Cyberlink PD issue not recognizing the technology.
we need to get a hold of an engineer.
|
|
|
i contacted customer service to see if i could get in touch with an engineer to resolve this issue, but it seems they think this is an nvidia problem....
here is my conversation (ATTACHED)
|
|
|
Quote
I think this discussion is getting away from the original post -- which is more about MSHybrid mode and Vidia Optimus. I don't think that OpenCL has anything to do with that.
OpenCL may work well for games, but I can't see any advantage to it over custom code for each GPU technology. As ointed out in the previous post, OpenCL just adding layers of additional processing slowing video rendering. Cyberlink should go back to true NVenc/CUDA support direct to the GPU card. I've already started looking for an editor that does.
GGrussell, Powerdirector does support CUDA and others, just not when paired with intigrated graphics. Premiere does support these technologies on latoptops with optimus, i tested it on the trial version. i don't want to switch to Adobe becuase of the extreme cost, so i hope Clyberlink can address this very soon.
|
|
|
Quote
This thread itself is The Petition! )))
Just add a post: "I am with you!"
I feel your pain... though CUDA support will gradually be replaced with OpenCL support over time, in fact yes, OpenCL uses CUDA interface to work with NVidia hardware, thus indeed, it might be faster. But not obviously: when I enable CUDA and use NVidia optimized effects, my performance is 10-15% less compared to Open CL (2 hardware platforms Intel +NVidia) plus CPU c++ code.
Vote for hope!
OpenCL has to work better on distcrete GPU's over intigrated ones right? in my case, not even OpenCL is using my nvidia card on my Surface Book, it's always using GPU-0 which is the intel HD620. Premiere is using the nvidia card by default, and it's butty smooth, but i don't want to pay $20 a month as a hobbiest.
|
|
|
Quote
Quote
We shouldn't have to go through all this and I complained about this issue several years ago so Cyberlink should be aware. They need to fix their software. This can also be an issue on desktops, but most desktop mobos have a way to only use the dGPU. No clue why most laptops can't.
Hi guys,
Just a small comment now as I am quite busy atm,
It seems that PD uses OpenCL interface to hardware, when available. You still need to trick it with NVidia profile to enable Optimus.
Once enabled, you typically have several OpenCL devices available:
-CPU device
-Intel Integrated graphics
-NVidia high performance card
Every device when connected to OpenCL interface is referred as "OpenCL platform".
And then, there is a generic openCL interface (IMHO) that connects to hardware specific platforms.
Each hardware platform is responsible for providing resources for calculation in hardware specific manner;
While generic OpenCL interface controls them all by means of unified OpenCL language.
It also provides control over hardware OpenCL devices, which in turn report bak their utilization and GFlops caps.
With this architecture, generic OpenCL provides task sharing between the hardware specific OpenCL platforms which in turn, assign the assined part of the job to their resources (cores and memory).
It also provides load balancing and profiling.
With this said, IMHO, True Velocity is just another name for that.
There is little Cyberlink can do about how OpenCL performs.
Note: not nesessarily this is the best option. For example, manually disabling software (CPU_ openCL platform on my system renders it more productive (15%) with replacing execution model to CPU c++ threaded code instead of OpenCL on the main processor; while two other platforms are still OpenCL controlled.
With OpenCL enabled on all three devices, none of them will peak up to 100. For example, CPU may only stay at 65-70%.
With OpenCL platform disabled on CPU, the CPU code peaks to 99-100%, and the other OpenCL devices are also more loaded (by 15-20%) compared to purely OpenCL configuration.
Finally, when you use OpenCL, real time performance anyway poor. Just imagine how many layers of abstraction, task splitting, controlling, balancing, asynchrous computing and profiling this baby adds. Hence, if you check the box, you may notice your preview is not continous or degrade or stops suddenly.
When you disable OpenCL, you benefit from Intel realtime sycronous processing (especially with Intl optimized effects), smooth previews but high render time...
That is why I encourage you to subscribe the petition to Cyberlink^
ADD SEPARATE OPENCL CHOICE FOR PREVIEW AND RENDER
So far Cyberlink ignores this...and I have to repeatedly check and uncheck switching between the modes.
My fortune:
- CPU at 99%
- Intel HD at 55%
- NVidia card at 45%
When rendering from 4K source to FHD target full color, full effect set (Intl optimized), some enchancing (levels) and packing that into MKV container with H.265 codec at high profile and quality at 11 Mbps.
So yes.. you can do that, if you want it badly.
P.S.: 2313 seems to be some 10% slower than 2258, still have to check on that.
VOTE!!!!
where is the petition?
CUDA seems like a way better option if we can get them to properly address optimus. i like PD16 a lot, but it's performance against Premiere is laskluster during editing, scrubbing, and effects. I bought a $2000 laptop to have a really good mobile editing experience, and at this point the software is athe bottleneck.
|
|
|
hello everyone,
i recently got into Powerdirector and have been using it on my Surface pro 4 and desktop which has an Nvidia gtx 1060. as my desktop is AMD Ryzen, it has no intigrated GPU, so it uses the Nvidia card and allows me to choose CUDA. the Surface pro 4 only allows intel synce or whatever it is. sinc ei like editing at my local coffee shop and brewery, i thought lets get something more powerful.
enter the Surface Book i7 w/gtx 1050
I think, wow this will be great, all the pwoer i need on the go. turns out, as this thread discovered, PD will not use the dGPU if there is and intel one available. i do not have the option to turn off the intel graphics, so need ot find a way to trick PD into using the Nvidia card.
we know PD is compatible with Nvidia, we just need to trick it, so i started looking through the registry, and found something called "GPUUtilityEx"which links to GPUUtilityEx.exe inside the cyberlink program folder. this utility seems to scan the computer and dictate the gpu in use. i tried altering the registry to indicate what my nvidia desktop showed in those fields, but whenever PD16 is loaded, those values get reset. so i removed GPUUtilityEx.exe from the folder, altered the registry again, and loaded PD16, and got some interesting results. under hardware acceleration options, it used to say OpenCl and i could select it, now it says "enable Intel Effect Accelerations/NVIDIA CUDA/AMD Accalerated Parallel Processing technology to speed up video effect preview/render" but the option is greyed out. so i look in the registry again and some of the values i changed remained, others went back to intel.
 https://photos.app.goo.gl/LSsUBW6QcalHFKyl2
https://photos.app.goo.gl/gAx9sSsN50uUClO03
so, we need to try and find the mechanism that PD16 has that selects the GPU. i was also able to delete the Nvidia Driver profile for Powerdirector and replace it with one that owuld allow e to select the GPU in the Nvidia settings, but that doesn't seem to work, power director is slecting it's own GPU, not being told where to go.
here are the registry entries from the Nvidia desktop
https://photos.app.goo.gl/w5b8opX7ubgaJY262
https://photos.app.goo.gl/BKljD94Iv257zA422
|
|
|
