See this article: https://www.geeks3d.com/20181220/how-to-enable-intel-opencl-support-on-windows-when-amd-radeon-graphics-driver-is-installed/ . The open cl is not affected with an amd cpu.

Quote We're talking about up to 30 FHD tracks playing simultaneously.

Smooth preview without lagging is one of my main concerns.

There's not a GPU and/or CPU combo with PD that will achieve continuous duration smooth FHD preview of 30 simultaneous H.264/H.265 FHD clips at say 24Mbps each.

OpenCL not of real value for playback decoding of unmodified clips.

Jeff

Quote

There's not a GPU and/or CPU combo with PD that will achieve continuous duration smooth FHD preview of 30 simultaneous H.264/H.265 FHD clips at say 24Mbps each.

OpenCL not of real value for playback decoding of unmodified clips.

Jeff

Thank you for your answer.
In that case, what would you suggest as a solution for such projects? (See youtube link above)

If it helps, I don't mind watching the preview in a lower resolution (as long as it plays smoothly without lagging), but as far as I understand it won't help much. The PC still needs to process 30 FHD tracks playing simultaneously, and only then downgrade the resolution for the preview window, right?


I still haven't bought the PC, so I'm open to any configuration.

Quote The PC still needs to process 30 FHD tracks playing simultaneously, and only then downgrade the resolution for the preview window, right?

Not if you enable shadow files. PD will pre-convert all your source clips to lower resolution versions and will use those when editing. When you produce the project, PD will use the full resolution (original) clips.

The only downside is that it may take hours to convert the clips, since PD treats this as a background process and there isn't any way to "force" it to prioritize this job. As each clip in the media library is processed, the little yellow icon on it will turn green, and you'll get the best editing performance if you wait for all clip icons to be fully processed before starting your editing session.

There is a another workflow where you create lower resolution versions of all your clips on your own ahead of time, using PD's batch produce feature, or a third party app like Handbrake or VirtualDub2, and use those files for editing. Then when you're ready to produce, you close PD and swap out the low-res versions in the source folder with the identically named original clips, so when you launch PD again, it has the full resolution clips to produce from.

Quote
Thank you for your answer.
In that case, what would you suggest as a solution for such projects? (See youtube link above)

If it helps, I don't mind watching the preview in a lower resolution (as long as it plays smoothly without lagging), but as far as I understand it won't help much. The PC still needs to process 30 FHD tracks playing simultaneously, and only then downgrade the resolution for the preview window, right?


I still haven't bought the PC, so I'm open to any configuration.

I'm not sure why one needs smooth without lagging playback during editing for those parts of the link that have 20-30 clips playing. If you need it for some reason, a few options might be:
1) Use PD render preview in those chunks that you have 20-30 clips simultaneously playing. A CPU bound slower process and needs to be repeated for every edit but when done will playback smooth during editing playback. It appears these chunks of 20-30 videos playing are fairly short duration.
2) Since they occupy say 1/30th of the finished frame size, reduce framesize from 1920x1080 to say 320x180 at a much lower bitrate, maybe 1-2Mbps. Obviously requires lots of preprocessing of clips which can be time consuming. This encoding could be done with CPU or GPU but significantly faster with any 4th, 5th or 6th generation NVENC GPU. Additionally, remove the audio channels in these playing clips if the production has one underlying audio track
3) Use item 2 for editing and then use unreduced framesize filenames for final production

Jeff

Quote

Not if you enable shadow files. PD will pre-convert all your source clips to lower resolution versions and will use those when editing. When you produce the project, PD will use the full resolution (original) clips.

The only downside is that it may take hours to convert the clips, since PD treats this as a background process and there isn't any way to "force" it to prioritize this job. As each clip in the media library is processed, the little yellow icon on it will turn green, and you'll get the best editing performance if you wait for all clip icons to be fully processed before starting your editing session.

Thank you. Very valuable info!
Importing all the tracks and letting them be shadowed before starting the editing - sounds like a solution for me.

So this brings me back to the hardware question:

Any suggestions about the Radeon RX 5700 XT vs Any of NVIDIA cards?
(The 5700 XT runs on PCIe4.0, but per my understanding doesn't support OpenCL, while NVIDIA cards do support OpenCL, but run on PCIe3.0)

This nvidia card should work: https://www.bestbuy.com/site/nvidia-titan-rtx-24gb-gddr6-pci-express-3-0-graphics-card/6320585.p?skuId=6320585&ref=212&loc=1&ref=212&loc=1&msclkid=b8ae6f93dad613d83bb2319ac67161e6&gclid=CLrLntLo2-YCFZarxQIdSFQMOg&gclsrc=ds .

... Or you could spend $2000 less and get an RTX 2070. This chart shows that it has the same NVENC capabilities as the Titan RTX, but with much less of the unneeded (and very expensive) super gaming graphics, which PD doesn't use

YouTube/optodata

DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°

....Or spend nearly nothing ($100) and get a used 10 series because even with lowest resolution shadowfiles, the direction he preferred, PD18 won't play 30 clips FHD simultaneously for extended duration fluently.

Jeff

Sure, but in the end the video will still need to be produced, so why not get a more capable/faster card?

YouTube/optodata

DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°

Quote Sure, but in the end the video will still need to be produced, so why not get a more capable/faster card?

More capable/faster, what makes you think that with PD18 release 2405? We are talking basic FHD and additionally for any significant Turing improvements to be realized, a current SDK needs to be linked.

To prove the point for you I put (25) 1920x1080 H.264 24Mbps source clips in the time line in a simple wall type setting to imitate the OP desired need, timeline pic attached.

I used hardware encoding and 2 different default produce options in line with the OP FHD need, MP4 H.264 1920x1080/60p 40Mbps and 1920x1080/30p 16Mbps with a RTX2070 and a GTX1070 both running 441.66 driver.

For MP4 H.264 1920x1080/60p 40Mbps produce, RTX2070 0.12% faster
For MP4 H.264 1920x1080/30p 16Mbps produce, RTX2070, 0.05% faster

That's a negligible difference, I'd put the extra ~$400 bucks into the CPU.

As a side note, I can't fluently playback the timeline either with shadowfiles at the lowest resolution (720x480) as was estimated so the OP goal of "Smooth preview" is not achieved.

Jeff

Quote That's a negligible difference, I'd put the extra ~$400 bucks into the CPU. As a side note, I can't fluently playback the timeline either with shadowfiles at the lowest resolution (720x480) as was estimated so the OP goal of "Smooth preview" is not achieved. Jeff

I don't have concurrent cards to test, so thank you for showing that there's no meaningful difference between those two nVidia cards. If CL ever does make full use of the Turing SDK, I would hope to see a bigger difference, but there's no reason to think it's coming anytime soon (if at all).

On the other hand, my i9-9900K/RTX 2070 combo is pretty close to being able to preview 25 FHD 60p clips all at once in FHD preview quality without using shadow files:



When the big music blast comes mid-clip, the scrubber is about 1/2 second behind the audio, but to me, that's still usable if you only need to preview/edit small chunks at a time. If I disable one video row of the tracks (20 remaining), the delay is almost entirely gone, and it should be possible to watch longer sections whenever there are 20 or fewer active tracks. The example video had many sections like that.

Here's the exact project I recorded, and I used an external hardware recorder because using CL's ScreenRecorder would have had a noticeable performance impact. It's definitely best to close all unneeded apps and services as my CPU was running at 100% @4.7GHz.

Here's my current CPU Passmark score for reference:



YouTube/optodata

DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°

Quote It's definitely best to close all unneeded apps and services as my CPU was running at 100% @4.7GHz.

One of the reasons I said for the OP to put the money from not buy your suggested RTX2070 into the CPU for his particular need.

I can play my 25 clip rendition about 20-30 sec and then significant issue with smooth preview without lagging in OP FHD need or a full 2+min in HD. I don't consider this meeting the OP bolded objective, the referenced sample video was nearly 5 min long and a desire for 30 clips. One would have to pause, play, pause to attempt to make it through playback at FHD. Hence my prior comment on the true need for smooth playback to edit this type of video.

I think the OP has been given ample information for their decision and goal of smooth preview without lagging of 30 simultaneous clips.

Jeff

optodata, your clip appears to be private as its showing a grey stipple in the forum page.

Jeff, Optodata, thank you both very much for the valuable info.

Jeff, I'll follow your advice, and invest more in the CPU. Probably will get the Ryzen 9 3900x or even the 3950x instead of the 3800x


About the graphic card - both of you were talking about NVIDIA cards and didn't mention the Radeon RX5700 XT. I suppose there's a good reason for that, so I'll get an NVIDIA card.

I will google around some more in order to see if there's still a reason in my case to invest in the RTX2070, but I will keep in mind your performance comparison between the GTX1070 and the RTX2070.
A little bummer about the unused PCIe4.0 bandwidth, but at least I will use it for the SSD.

Thanks again, guys

Quote About the graphic card - both of you were talking about NVIDIA cards and didn't mention the Radeon RX5700 XT. I suppose there's a good reason for that, so I'll get an NVIDIA card.

Both brands have issues with PD that appear to go unfixed for years but I've had better luck with Nvidia with respect to my workflows. Not to paint with too big of brush but PD and AMD VCE couldn't achieve proper bitrate encodes, particularly H.265, and was slower at comparative price point. Since your interest is basic FHD, probably not a big factor. This is a PD encoding view utilizing the SIP core, not gaming or the GPU in general.

A few years back CL pushed a patch that said they incorporated support for all AMD GPUs with VCE3.0, turns out more work was required and CL withdrew the supporting statement, discussed in a forum post if you search. Been pretty quiet for ~3yrs but now the 2405 release states "Adds support for enabling AMD VCN2 hardware acceleration." VCN2.0 replaced VCE4.0 in mid 2019. I brought home a compliant VCN2 Navi 10 AMD GPU but haven't evaluated for my needs in PD18 yet.

Jeff

I hadn't finished filling out the YT questionaire when I posted, but now the video is accessable.

YouTube/optodata

DS365 | Win11 Pro | Ryzen 9 3950X | RTX 4070 Ti | 32GB RAM | 10TB SSDs | 5K+4K HDR monitors

Canon Vixia GX10 (4K 60p) | HF G30 (HD 60p) | Yi Action+ 4K | 360Fly 4K 360°