Hello, and thank you for reading. This post describes several issues (both bugs and performance lower than expected) that I have discovered so far in using VP22. It's a long (and very thorough) post, but my hope is that I can come out of this with better future results, and the devs can hopefully glean some knowledge using a real-world example and make some improvements.
Disclaimer. These are my tests and my results with documentation. I hope/believe that there could easily be something I am missing and if I sound sad it's because I was hoping for BIG gains using the new VP22 engine. I am not here to trash VP22 or VP in general. I am here to provide tested feedback and to learn what I could do better on in settings.
First. the scenario.
COMPUTER SPECS (consistent throughout ALL tests referenced in this post):
CPU: Intel core i7 4790K @ 4.0 GHZ
GPU: ASUS RTX 4070 Ti SUPER w 16 GB VRAM on latest NVIDIA Studio driver
16 GB system RAM
Windows 10
Project files (original OOC and proxy files) on a very new Crucial SSD, rendered to an internal NVME SSD (not same SSD as program install or Windows install)
QSV is off in settings (and renders were all HEVC using NVENC, no QSV) to make sure that I lean on the GPU (strongest part of my system) and can be sure I am not "confusing" anything. Trying to isolate GPU performance as an aid to Vegas.
For context, this project was filmed in 10-bit HEVC. Filmed in 6K "open gate" (3:2) on a Panasonic S5iiX using VLog at 29.976 fps.
The following details about the edit are provided for analysis by devs and other users to help with troubleshooting/performance improvements, but I am NOT looking to fight about whether my methods for the edit are how everyone else would do it in terms of STYLE (tips are of course welcome but the real point here is to show what I have discovered about VP22 vs VP 19 in terms of performance and bugs and can show empirically).
The edit:
For edit, I cropped to from the original 3:2 open gate down to a 17:9 image (to get full side to side stage coverage) using "Pan/Crop," added a lightly-applied Sharpen (because even at stopped down and high res, far away people in a high-detail scene benefitted from it), and had two identical video tracks but one was stacked on top of the other, very slightly offset horizontally by a couple pixels with some Bezier masking added (why is a long story). I knew this would result in a 16:9 render with very slight letterboxing on the top and bottom, and this was intentional. I placed a solid color (black) underneath the video tracks in order to totally ensure that my crops (in case I decided to reframe later, which I never ended up wanting or needing to do as I remained happy with the crop result) would always have the same consistent black borders on top and bottom. Whether or not this was strictly necessary is not really relevant, but I am posting this fact for FULL disclosure to aid in analysis of results. There were also many text items on the very top track of the project to demarcate sections of the show. Simple text, nothing fancy, no effects on the text.
In order to achieve the desired look but save on processing power, I went into the Color Grading panel on the original source track, used an input LUT designed for VLog cameras, and added a store-bought Look LUT on the back end as well. Once I got that pretty much how I wanted the project to look, I exported those settings as a Custom "Look LUT," then turned off color grading on the source tracks, then applied my newly made LUT (saved of course in .cube format) to the Video Bus only. Last step in this regard after much time staring at it editing cuts/transitions (but before rendering) was to also add a slight curves adjustment to the Master Video track (or bus, however you want to call it). So to be clear and to sum up, when editing and rendering:
- NO Color Grading was on the two source tracks (remember I turned it all off on the source track(s) after creating the LUT I applied to the master), only sharpen (and some Bezier masking on the top copy in addition to it being shifted horizontally by a couple pixels, to mask a couple teeny tiny flashy dots in a couple places caused by venue lighting).
The render(s):
- Renders (in both VP19 and VP22) were done using the HEVC NVENC template, at 3840 x 2160 @29.976 fps, "high quality" profile, full-resolution render quality "Best," Rec.709, Limited, VBR highest 40,000,000, average 30,000,000, 320 kbps AAC for audio.
I always wanted to use NVENC either way to (theoretically) leverage what SHOULD be the massive power of this RTX 4070 Ti Super. I honestly would rather render to h264 AVC (still using NVENC), BUT the only template Vegas has that puts out 10-bit is HEVC, and I also didn't want the file size to be completely ridiculous (it is a LONG show that is going to end up on Vimeo). The full show at these rates will be about a 27 HOUR render. I can only imagine what it would be if using H264.
After having previously (last week) done several versions of the exact same material/ same render settings on VP19 in order to test the effects of different options like "legacy GPU rendering" on/off, and then (today)hoping to achieve faster renders with my RTX 4070 Ti Super, but with the hoped-for benefits of VP22 as described (one of the MAIN reasons I upgraded to VP22 was the described engine/render/playback improvements, particularly for NVENC renders and for editing playback of HEVC, which this project was both shot and rendered in), and screenshotting the results (performance, specs like render time, CPU usage, memory usage, average FPS, render times, etc) today after installing VP22, I did the following:
- made a copy of the project to test with VP 22, and got all my program settings for VP22 to match how they were in VP 19
- I rendered the same exact material with the exact same render settings in VP 22 with the intent of seeing whether Vegas Pro 22 really is significantly faster as advertised due to the new engine.
Here are my conclusions, and ALSO a couple bugs/performance issues I have noticed about VP22 as compared to VP19.
Conclusion 1: no noticeable increase in playback performance while editing. It's that simple. It still struggles. It struggles at Preview quality: half and full, and it struggles HARD when preview is set to Good (half or full). Just like it did in VP 19. I know that 10-bit HEVC is a lot to ask, but with a RTX 4070 Ti Super (yes I know my CPU is a little dated but not THAT bad), i figured at the very least that VP 22 would struggle less than VP19, being that is supposed to be a strong point of VP22's new engine/decoders.. It was the same. Like, the exact same. If there was any improvement it is not detectable. Color grading off...still struggles. In VP22 just like in VP19.
I have "enable experimental HEVC" on in VP22 (is that a mistake?), and I of course have the 4070 selected for GPU rendering, optimization on, all that. And in my NVIDIA control panel, it is set up correctly there too. And I even went into Task Manager (in both my VP19 and VP 22 tests) and made sure Vegas was set to High Priority, and had no other apps running during any of these tests.
Conclusion 2: VP 22 is only very slightly better than VP19 when rendering HEVC using NVENC with all settings the same.
All tests were renders of a 2 minute 6 second clip of the exact same thing with the exact same settings.
Results of VP 19 test, only difference between these two renders depicted in the picture (both in VP 19) being trying legacy HEVC on and then off:
https://drive.google.com/file/d/1QpMyuw_IIBVXFNl3D8lH0Vn-lxJi-cG1/view?usp=sharing
Result of the same exact material and render setting in VP22 (but in VP22 there doesn't appear to be anything to try in regards to "legacy HEVC, however, I did have "enable experimental HEVC" on):
https://drive.google.com/file/d/1EZHz26Zuwznm7oYYBoJBqOrSBBveX3_i/view?usp=sharing
As you can see, the render time is very very similar. Not exactly the big speed upgrade I hoped for with the new engine. About the only benefit I can see from this test (and to be fair that slight reduction in render time might be solely attributable to the fact that I accidentally rendered a selection that was one second shorter at 2 minutes 5 seconds of material as opposed to the 2 minutes 6 seconds in VP 19 version) is that it was about 0.25 to 0.5 fps faster average. But in all cases it was around 2-3 fps average speed.
Also note: in VP19, Legacy HEVC off used slightly more CPU but way less disk usage (no idea why) than legacy HEVC on. Legacy ON was roughly 2-3 minutes faster but still took about 10 times as long to render as the length of the file.
And note that VP 22 did not fare much better in render time...slightly better than VP19 with "legacy HEVC" on, by about 2 minutes 20 seconds total improvement, but not a huge upgrade. Still over 20 minutes to render a 2 minute file.
Here is the Media Info showing, and I guess, comparing (although they are basically identical files, with one BIG caveat discussed below in another conclusion, see my immediate reply post about "conclusion 5") the file outputs resulting from rendering in VP19 (with legacy HEVC ON), and rendering in VP22 (experimental HEVC enabled):
https://drive.google.com/file/d/1TLFODT8oNiq9WK4WTAQf5O8lXgUC4b6x/view?usp=sharing
Conclusion 3 (bug report): VP22 has a bug in preview (NOT present in VP19), but will thankfully render without the bug.
See links just below this conclusion for evidence. I show it happening in VP22, but NOT happening in VP 19.
In VP 19, my 17:9 crop (in edit) shows properly at all levels of preview, and it renders as expected... 16:9 file with slight letterbox on top and bottom as intended/expected. In VP 22, the preview, IF set to "Preview" quality, distorts the image. Maybe distort is not the right word, but as you can see, it displays the content with SIDE letterboxes and fills from top to bottom (not correct!), but I found that when you change the preview quality to "Good" it thankfully fixes itself (reminder, there IS supposed to be slight letterboxing on top and bottom because the crop is 17:9 but the project/render are both 16:9)... whereas in in VP 19, it just looks right the whole time in all preview qualities.
VP19, no bug
https://drive.google.com/file/d/1X7HWzghwnmrBeARFC-ER1aR4NZSb611W/view?usp=drive_link
VP 22, bugged (watch what happens when preview quality is changed back and forth between "Preview" and "Good")
https://drive.google.com/file/d/1q8IqpOPflSeQYiesGwFDU5ZHPFwrfzKl/view?usp=sharing
Scary, but at least it does work out ok if you export a cap, and also ok when you render.
Conclusion 4 (possible bug?): VP22 lost the CG settings from VP19 (copy of same project, same drive, same files) and I had to manually redo them
Reminder for context: I saved my Color Grading settings as a Look LUT (in both versions) and also had the identical curves setting in both versions.
When I made a copy of my VP 19 project file so that I could test same in VP22, stored on the same SSD and using all the same source files, I was excited that it imported into VP22 fine and at first everything seemed great. But then I noticed two things: the preview bug noted in conclusion 3, and that for some reason, the VP22 version did not have the Look LUT in it/applied to it on the Master Video track/bus/whatever you want to call it.
I was able to go get the Look LUT I created from my PC (same location as VP19 project) and manually add it into the VP22 version the same way in the same place, but I was scared for a moment when I looked at the screen and saw that the color grading from the VP19 version was initially not present in the VP22 version. Good thing I had saved it as a Look LUT (.cube) and had not just had it as a manual group of settings, cause replicating all that by hand would have been rough.
The result was not a game breaker cause I was able to fix it, but I hope this is useful knowledge to the devs and to any users who may experience this as well.
THANK YOU so much to anyone who made it through this post. I repeat again that I am still a big fan of Vegas, and still intend to use it. And I hope this commentary is helpful! I will be happy to provide further info to any questions to the extent possible and have a few more screenshots if even more detail is need. Devs, please feel free to reply or message me if any of this is of interest to you for further discussion. Users, I welcome your comments/questions.