Okay, I could get it to work, but had to kluge.

And Moderator? If you're reading, I think I might know what the problem is with PD here, why users are locked out of using Hardware Acceleration when an Intel iGPU is detected.

I think it's a simple UI problem. Probably from long ago befor Nvidia encoding came into use.

I think that the UI doesn't have any room to show the "Hardware Acceleration" button. Either that, or a programmer didn't offer it once the Intel is detected., and a button for "Hardware Encoding" is not being painted properly.

Here's what I did:

I disconnected my second monitor that is driven by the Intel GPU.

Then I started Win10 and PD 18.

I ran the Optimizer again. Nothing changed on that screen. (That is, it did NOT say "Yay! You have Hardware Acceleration for Encoding.")

But when I went to the Produce page, now it showed "Hardware Acceleration" WHERE IT USED TO SHOW QuickSync.

I think that's a clue.

Now, remember, when I had both GPU's pushing montiors (making both GPU's active), and after I used Win10's Graphics Settings feature to point PD 18 to my Nvidia, the "Use QuickSync" hardware encode option STILL showed in the GUI. But it was NOT selectable. So the "Use Quick Sync" button is kind of a default in the code.

I suspect that the UI is misprogrammed, showing me a useless button.

Therefore, I think that when PD sniffs GPU's and sees an Intel GPU, it always paints "Use QuickSync" in the UI, despite the fact that the button cannot be selected. (The UI might also be painting the "Hardware Acceleraton" button. But if it is, it's off screen somewhere and cannot be selected.)

Moderator, might I suggest that you ask Engineering to look into the UI code for the Produce page, and see if a flag is preventing "Hardware Encoding" from showing when an Intel GPU is flagged first? Or if the "Hardware Acceleration" button is being called in the code, whether it is formatted properly to display properly in the UI?

If someone added the code for the Hardware Acceleration button after QuickSync was already implimented (as I suspect is the case), then they might not have thought about the case of having two GPU's in a computer. That is, this bug might be a simple overisight in the UI code.

This has been a particularly vexing problem for PD. None of my other video editors have this problem when both an Intel iGPU and an external GPU coexist.

BTW, the Nvidia Hardware Encoding was at least twice as fast as my UHD 630 for an easy 420p video.

Disable in BIOS the intel GPU. Probably now is set on "Auto".

All the settings in the software that would seem to associate nvidia to the CL software actually associate only the GPU cores of that GPU, they don't "touch" the separate ASIC's that are dedicated for encoding/decoding video. So those are asigned on "first come first served" basis...

IMO that's an oversight in the drivers from the GPU manufacturers. They care only about the gaming cores...

In my Dell BIOS, there is no way to disable the UHD. It says on the BIOS screen that if you select other than the Intel, that it is "still available" to the OS. Perhaps a limitation of Dell and their BIOS's, which don't allow for overclocking memory either.

As for being a driver "first come, first served" problem - I don't know enough to know. But I do know that a "Magic" video editor that I have works fine with both drivers on my system, and the editor sees whichever GPU is selected via Win10's Graphics settings.

Update: And, in fact, even PD 18 is seeing which GPU that Win10 is point to it, because when I select my Nvidia for PD 18 in Graphics Settings, I can no longer use the Intel GPU. So I'm thinking the driver switch is taking place correctly. This is why I think the problem lies elsewhere.

So I think it's worth having a CL Engineer look at the code for the GUI where "Use QuickSync" vs. "Use Hardware Acceleration" paints on screen. I'm thinking that a conditional branch or a screen write is mis-coded. Easy enough to check, and not much lost if I'm wrong. (But a BIG win if this is the (simple) problem.)

BTW. on another note, Audio Waveforms paint much faster in PD 18. Looks like that problem is fixed?

Here is a recent recent tutorial on getting the GPU to work: https://youtu.be/vjaDd0TIth4 . Let us know if any of this help.

Quote In my Dell BIOS, there is no way to disable the UHD.

Being a desktop you can just disable it in Windows. On most of the laptops, the screen is physically connected to the Intel output, but here is not the case.
Device Manager, right click on the Intel display adapter and choose "disable".

Hi Gentlemen:

Thanks for the suggestions.

I had done everything in the video.

But SoNic67's idea of disabling the iGPU from Device Manager is easier than unplugging my 2nd monitor from the Intel GPU and rebooting. (Until CL fixes this problem, that is!)

You should be able to connect both of your monitors to the Nvidia card which supports up to four monitors. Leave the MB graphics connection disconnected. The Nvidia card should then be fully enabled.

Other Dell users have Nvidia cards and they sell them at the Dell websites for their pc's. Maybe a chat with their support staff can help on what you need to do.

Tnx. Yeah, I know I can do that. But I have reasons for having the 2nd monitor driven by the UHD. (OBS Studio.)

Quote
But SoNic67's idea of disabling the iGPU from Device Manager is easier than unplugging my 2nd monitor from the Intel GPU and rebooting. (Until CL fixes this problem, that is!)

So did that work? Disable the Intel GPU while editing in PD and then enable back?

I don't think CL will "fix" this, becaus ethey rely on GPU drivers for that. As of now, there is no mechanism in drivers of asigning the dec/enc ASICs to certain apps.
Only if CL would develop their own encoding software that would use OpenCL/CUDA for encoding and not the free ASIC, then they could do that.
They could also get permission from nVidia to include the old CUDA encder dll's in their software distribution, but somehow I doubt nVidia would do that.

Bandicam suggests a grey method - users to download those and install separatelly:
https://www.bandicam.com/support/tips/nvidia-cuda/

Yes, your trick works. Of course, I lose my 2nd Monitor when I disable its Adapter (so I can't use the Mulitmonitor layout for PD).

And all my open windows move to my primary, and I have to move them back after I re-enable the UHD Display Adpater.

Someone in this forum (or in a competitor's forum) uses an HDMI switcher for his 2nd Monitor, so that he sometimes run both his monotors off the Nvida.

Quote Yes, your trick works. Of course, I lose my 2nd Monitor when I disable its Adapter (so I can't use the Mulitmonitor layout for PD).

And all my open windows move to my primary, and I have to move them back after I re-enable the UHD Display Adpater.

Someone in this forum (or in a competitor's forum) uses an HDMI switcher for his 2nd Monitor, so that he sometimes run both his monotors off the Nvida.

I have 2 HDMI switches and can run either monitor off my nVidia or iGPU card, and I've made sure that all configurations have the same profile under Win10 so I can switch between them at will.

I also use a very helpful 3rd party app called DisplayFusion, which does great things like managing taskbar icons on a per-monitor basis and remembering exact window sizes and positions, even when programs (ahem, PD) don't offer that functionality.

Quote Someone in this forum (or in a competitor's forum) uses an HDMI switcher for his 2nd Monitor, so that he sometimes run both his monotors off the Nvida.

Using switchbox approach has been around for years, https://forum.cyberlink.com/forum/posts/list/49546.page#post_box_260582 I'd agree with Sonic67, I don't think CL will "fix" anything as I don't think anything is really broken. For a desktop, I think they simply detect primary display and use that device, it's really that simple. I also really doubt CL will develop their own CUDA based encoder, if anything they'd license from some of the many out there but that's $'s so I don't think that will happen in this price market either.

CL removed the old API interface so old CUDA dll's and new PD versions won't work. That was seen in the forums when Nvidia stopped the encode bundle and CL PD users had no options until a patch finally came along. Nvidia gave developers several sunset years notice so it's not like it was a surprise.

Jeff

Quote
For a desktop, I think they simply detect primary display and use that device, it's really that simple.
Jeff

Okay. Just to be clear, my Nvidia IS my primary display.

When I have Graphics Settings pointing PD to my Nvidia, I still see the option to use QuickSync for Hardware Encoding, even tho PD shouldn't be seeing the Intel iGPU. Even though the QS shows as a choice, it is not selectable.

That's why I believe that PD is confused somewhere.

As one more additional thought: I use OBS Studio 25.x for screen recording. The OBS folks collaborated with Nvidia for version 24, and use Nvidia's NVENC.

When I set Graphics Settings to point OBS to my Nvidia, then I can do hardware (NVENC) encoding in OBS.

When I set Graphics Settings to point OBS to my UHD, then I am locked out of using NVENC in OBS.

Quote Okay. Just to be clear, my Nvidia IS my primary display.

When I have Graphics Settings pointing PD to my Nvidia, I still see the option to use QuickSync for Hardware Encoding, even tho PD shouldn't be seeing the Intel iGPU. Even though the QS shows as a choice, it is not selectable.

That's why I believe that PD is confused somewhere.

Sounds like you are describing a windows display setting, that is way downstream of primary graphics adapter being defined. It's controlled during POST. If you have BIOS with graphics control and it's set to auto, what can happen during POST is sometimes you can get iGPU, sometimes Nvidia adapter, it depends on cabling, type of cable, type of connection, as one is talking ms to establish initial handshake. Some desktop BIOS chipsets don't give you control and routes everything initially to the iGPU (which will then be primary adapter) and then the PEG/PCI as secondary, in such cases disabling the iGPU via the device manager as was suggested is an option.

Jeff

I understand.

In my BIOS I have the Nvida pointing to my primary monitor as a hard set.

(If I tell the Dell BIOS to use "Auto," then it is as you say. The BIOS starts off with my secondary monitor as my Primary, and I see the Dell Boot screen on my secondary monitor. Then after WIndows starts - downstream, as you say - Windows reverses things, and my primary monitor becomes my primary.)

I created a Support Ticket for this.

(At least I think it's a Support Ticket. It says it's a Question, CS002140208.)

For anyone interested, here is a variation of a hastily made, 5 minute video I sent to CL to describe the problem

https://jmp.sh/6an8KUC

Good news/bad news: I received a relatively timely response to my Support Ticket.

And I can see that someone at CL watched my video.

That's the good news.

But apparently they did not understand the problem and/or think that I am an idiot, because Tech Support said that I should follow a FAQ from PD 10 days about clicking "Hardware Acceleration" to enable Nvida Hardware Acceleration.

Except for the small fact that "Hardware Acceleration" is not a choice in the Produce page when an Intel iGPU exists in the system.
As I clearly showed in my video.

Sigh.

I asked the Tech Support person to move this up to Engineering. Or if CL is so sure this is a simple problem, then I asked CL to make a video showing us how to select [Nvidia] Hardware Acceleration in a system with both Intel iGPU and Nvidia GPU.

Another update. (This thread is morphing into griping about Tech Support.)

I got a 2nd response from Tech Support. This time they wanted me to send proof of purchase.

Why? What difference could this possibly make? I made a video of PD 18 running on my computer. I showed the version number. So we know it really is PD18.

I showed a SN in the video I made for CL. I gave CL my Activation Key when I submitted my Ticket. Suppose I had driven to JL_JL's house and made a video of this problem occurring on his licensed version of PD18? It doesn't change the fact that the problem exists, no matter who paid for the license.

I don't believe that Tech Support has sent my video to Engineering.

This is getting silly.

(Moderator: I don't know what kind of computer you have. But if you have one with an Intel iGPU and an Nvidia GTX, perhaps you can replicate my results and make your own report to CL? They might listen to you.)