Premiere vs Vegas - 4K video decode/preview video cross track test

LukasDr wrote on 7/25/2021, 4:25 AM

I have access to Premiere for few days and notice huge difference when it comes to performance of decoding and video playback in general. To test I used 3 clips - 4K 100Mbit, HEVC 8 bit. Resample disabled in project.

Test includes 3 tracks with 3 clips. All clips are transparent to force decoding and mixing them at the same time. Vegas overloads NVDEC decoder and it's impossible to playback 3 clips at the same time while Premiere goes smooth. Also Premiere does not have any hiccups, stutters, lags when start playback from random places. Vegas has it very often, and need 2 or 3 repetition from the same place to get smooth playback.

Please look at video here:

Test project to download from here: https://1drv.ms/u/s!ArT6bvS-V574l6V4kLAqseEF4muQVA?e=QtmqHF

Any idea why it happens? Any way to make it work better?

Lukas

Comments

RogerS wrote on 7/25/2021, 4:28 AM

What's your CPU? Can you test the Intel QSV decoder too?

LukasDr wrote on 7/25/2021, 4:43 AM

Hi, all is in my signature. Ryzen 5900X + RTX2060. On Intel Quick Sync is the same issue. But didn't test Premiere on Intel.

Premiere has lower CPU usage, lower GPU decoder usage, higher performance.

Vegas has higer CPU usage, higher GPU usage, lower performance. only 2 clips were able to go smooth, while Premiere - 3 clips.

Musicvid wrote on 7/25/2021, 5:30 AM

Premiere does a background render, Vegas does not. There are disadvantages to either scheme.

LukasDr wrote on 7/25/2021, 6:31 AM

Interesting. Could you say more about background render? What can be worse than suttering video playback?

I see nothing working in background when not play. Computer just is idle.

JN- wrote on 7/25/2021, 7:18 AM

@LukasDr It's known for some time, plenty of mentions in the forum of slow playback/render issues with NVenc compared to say Intel.

It would be useful if you supplied the same or similar project as a complete download project. That way other users (with different specs to your PC) here could give VP feedback.

At the moment we only have your PC by way of comparison. I don't know anything about Premier but it's very important to compare using similar conditions.

In VP playback can be enhanced using proxies, and also using the "Adjust Size for optimal playback" item by right clicking the preview screen.

So you need to make sure, ideally, that all playback enhancements are OFF in both software.

"Could you say more about background render?" This indicates to me that you may not know enough about this, I certainly don't with respect to Premier.

If you care to take a look at the first post in the Benchmarking thread, link is via my signature, you can see that I have requested that users apply a fairly rigorous process especially for testing playback.

Thanks for taking the time to do the comparison, it's useful.

RogerS wrote on 7/25/2021, 7:34 AM

Ryzen doesn't have Intel QSV so not possible to test that with this system.

JN- wrote on 7/25/2021, 8:00 AM

@LukasDr These links below give somewhat similar suggestions for improved playback performance in Premier. So a good starting point in Premier is to make sure that most of them aren’t enabled. Of course you still need to have HW Acc. enabled, but not using artificial low resolution proxies, not disabling FX etc.

In VP the suggestions in the Benchmarking project will take care of most artificial playback enhancements.

https://techshare.in/9-ways-to-fix-adobe-premiere-pro-playback-lag/

https://fstoppers.com/originals/six-ways-improve-timeline-playback-adobe-premiere-pro-cc-239017

https://filmora.wondershare.com/adobe-premiere/premiere-pro-playback-lag.html?gclid=EAIaIQobChMIw5SaiqH-8QIVGO7tCh3TQw2gEAMYAiAAEgJ-8vD_BwE

https://motionarray.com/learn/premiere-pro/premiere-pro-playback-lag/

LukasDr wrote on 7/25/2021, 1:10 PM

Hi @JN- Thanks fot detailed description. I'm fully aware that both software should be compared in the same condition. So this is as I have done it. Both Vegas and Premiere does not use proxy, have the same project settings, same playback resolution (1/2 = 1080p to see entire frame), no prerendered files. It's just super simple project with imported 3 clips on timeline with transparency. I tested the same scenario on my Intel laptop. Vegas behave exactly the same. Difference to desktop RTX is that Intel Quick Sync can handle just one clip, when second appears on time then gets lag. I couldn't test Premiere on laptop. But I predict the same behavior.

Here I uploaded test projects for Vegas and Premiere: https://1drv.ms/u/s!ArT6bvS-V574l6V4kLAqseEF4muQVA?e=QtmqHF

If anyone have time to download and do the test on his machine AMD/AMD, Intel/Nvidia, AMD/Nvidia. Thanks.

Everyone who run both software will see where is difference. The most visible is stability of framerate in Premiere. There you can be sure that wherever you put cursor and play, the playback will be smooth. Hmm.

Lukas

JN- wrote on 7/25/2021, 2:06 PM

@LukasDr “I'm fully aware that both software should be compared in the same condition”

I figured that, but no harm in confirming, without that the testing would be meaningless.

I'll have a go at it later in VP18 and post my experience here.

Why have the playback at half, is it just to put extra stress on the system for the purpose of the test?

JN- wrote on 7/25/2021, 3:06 PM

@LukasDr

Laptop using Nvidia laptop RTX 3080 used for HW Acc and File I/O HW decoding ... 40, 4, 2.5 fps

PC using Nvidia for HW Acc. and also using Nvidia for File I/O HW decoding ... Full, 40, 2.5.

-----------------------------------------------------------------------------------------------------------------------------

PC only below, see signature for specs.

HW Acc and File I/O decoding left to right below ...

#1 If I used Intel Intel I got Full, 35, 15

#2 If I used Intel Nvenc I got Full, 40, 4.0

#3 If I used Nvidia Intel I got Full, Full, 35

So, this is what I would expect to get on my PC. The #3 item is optimal using Intel for File I/O HW decoding and the Nvidia card for HW Acceleration. The Intel iGPU is better optimised for decoding by VP than the more powerful discrete GPU, but we all know that by now.

Note that all figures above except "Full" fps are approximate. They obviously don't match your "Full" premier playback experience.

"Adjust size and quality for optimal playback was unchecked" and no proxies used

LukasDr wrote on 7/26/2021, 4:50 AM

Hi @JN-

Thank you very much for tests, very informative. General conclusion is that Vegas does not utilize hardware's theoretical power. On any (Intel nor Nvidia) GPU it does not work smooth as it could. Premiere shows that it's possible on the same hardware and should work fine with three 4k HEVC clips played at the same time with no lags in any case from every place.

I have one interesting observation - the stutter appears mostly on cuts of clips. When it stutters decoder peaking 100% - which is does not make sense. But when start from place where two clips are on timeline it sometimes goes smooth and decoder usage has only 2x load related to single clip - around 60%. It seems to suggest that there is some bad optimization or bug. On Premiere I didn't notice anything like that. Every single playback start smooth if decoder isn't 100% used. And from my test come conclusion that both decoders (Nvenc and IQS) should process three 4k clips at the same time with no issue.

Developers definitely should work on optimizations on this area. Stutters and lags are very often here.

Look how it works on my Intel laptop.

Lukas

RogerS wrote on 7/26/2021, 4:54 AM

Maybe test again with VP 19 soon.

LukasDr wrote on 7/26/2021, 5:36 AM

Yep, for sure. I will be testing V19 and will see (anyone knows when it will be released?). For me Vegas is way more intuitive in basics than Premiere. But the most disturbing is performance and stuttering video preview despite hardware power reserve. It's been like this as far as I remember. Trying Premiere just surprised me, how smooth it working.

JN- wrote on 7/26/2021, 5:45 AM

I’ll test again using VP19 when it lands for sure using your project and give my feedback here. Personally I'm as interested in seeing if there is any fix for the lack of Nvidia Canon R5 decoding codec support. See my comments here. https://www.vegascreativesoftware.info/us/forum/vegas-pro-18-update-5-build-527--129048/

It’s possible that if one is fixed then the other may be also.

RogerS wrote on 7/26/2021, 6:57 AM

I have also seen the stutters and lags in getting up to full speed you describe. I think it's related to the modern so4compound decoder. As each version has improved upon it since 15 I am hopeful that 19 will be better.

When will it be released? Last year 18 came out in August, so soon?

Dexcon wrote on 7/26/2021, 7:30 AM

The current 18/19 offer ends 1 August. Last year, the 17/18 offer ended on 2 August - and VP18 was released on 4 August (my AU time-zone). Following that scenario, VP19's release should be 5 August or not much later. Just an educated guess - no inside info.

vkmast wrote on 7/26/2021, 7:48 AM

@Dexcon seems that offer is now valid until August 3, 2021. I checked the /us/ and /int/ urls.

Dexcon wrote on 7/26/2021, 8:40 AM

Thanks @vkmast ... and thanks for updating the Vegas Price Watch thread (where I got the 1 Aug date from earlier). Two whole days extension of the offer ... that's marketing silliness. I honestly hate to say it, but this makes Vegas Creative Software (aka MAGIX) look like that they have no idea about what they are doing and experiencing a lack of strategic planning - it looks like its being done on the run.

LukasDr wrote on 7/26/2021, 8:46 AM

Any info about feature list?

Dexcon wrote on 7/26/2021, 8:50 AM

This is all so far:

Sneak Peek of the new version

Adjustments tracks

Improved format support

Range-limited color wheels in Color Grading panel

and so much more....

... and https://www.vegascreativesoftware.info/us/forum/vegas-pro-19-first-look-in-commercial--130620/?page=1

Howard-Vigorita wrote on 7/26/2021, 11:44 AM

@LukasDr I sometimes use transparency like you did when shooting 4k with a single camera and want to get a multicam-like effect by duplicating the camera track and zooming it differently using pip fx. The resolution loss of the blowup looks fine as a transparent overlay as opposed to the cut over to another camera I'd do instead with multiple cameras. But I avoid the double frame rate you used because it's so back breaking. Unless I'm shooting something with allot of movement that I want to do stop or slomo with. In which case I usually go for 120 fps up to my camera's max of 160.

But it looks like an interesting test... I'll run it on my machine which has both Nvidia and AMD gpus later today soon as it's available. And on my faster machine soon as I hear back from Magix sales and work out a way to take advantage of that 18+19 deal... don't have enough 18's to go around at the moment. Btw, did not see any hevc clips in your rar... just avc. Be happy to test that too as that's what I usually shoot.

eikira wrote on 7/27/2021, 2:10 PM

i will state this. vegas has in general really troubles with transparency. i see that often with anything that use feather. for example if you enable feathering in masking stuff, vegas is really bad with that, no matter how simple the mask is, it can be only a square of something small with feather enabled it starts to stutter and slow playback. had that in the past with my intel setup, still have it now with amd computers. in my eyes, it has nothing to do with gpu or cpu performance in general, but with how transparency is handled in vegas.

LukasDr wrote on 7/27/2021, 2:40 PM

@eikira yep, it can be as you said, but notice that in my case when it stutters decoder peaking 100% It's nonsense because decoder has enough power to decode. It's some bug. Because sometimes it not overloads decoder and mostly overloads. It's not 100% predictable.. It seems to be more complex issue. I have created support ticket. Perhaps devs will notice that issue and do further improvements.

Howard-Vigorita wrote on 7/27/2021, 3:02 PM

I do see what appears to be a feathering issue in one of the fx I use. But not with transparency the way I usually do it with track composite envelopes. I use composite envelopes both for watermarking entire videos with a logo and to overlay one video track with another. Like doing it with envelopes so I can fade watermarks and video overlays in and out without splits.

See no show-stopping performance issues doing that. But I do see massive slowdowns in the Vegas text media generator. At first I thought it was the rolling glow effect but it turned out to be the drop-shadow slowing every render to a crawl. Only use this cumbersome effect as a recurring opening so I work around the issue by pre-rendering it to an intermediate.

ps: here's an example of what was supposed to be a test clip I did for my 1st experiment with a 4k, 1-camera shoot.

All the zooms and overlays done as described above with the 1st release of v18 using a dup of the camera track shot hevc 2160p/29.97 at 60 mbps. Rendered to 1440p and thought the digital zoom/cuts and overlays looked reasonably decent. Went so well I've been shooting with the same camera, same settings ever since. But usually multicam without digital zooms. I did later upgrade the lens from the vintage Canon 24-70 zoom that I borrowed off of my still camera for the test-shoot. To the newer IS version that works much better with heavy-handed piano players on the wooden floors of my living room.