Thanks RogerS I moved the AI Torture test results to the proper thread.

@RogerS Sorry, I screwed up the time I just entered for VCE render time for the 11900k/6900xt... should be 02:04 instead of 00:38.

Hi Howard, I just fixed it, please check. Interesting to see one system of your faster overall and the other way faster for AI.

I hate to admit I'm a little lost in this thread, so apologies if the answer is obvious here, but - if I were to encode this project on multiple GPUs (same host PC) for comparison, which project scene should I be using? Is it this one: "VEGAS Pro Fanmade Ad - The All New VEGAS Pro (Update 1.01)"?

Which encoder should ideally be used? Just the relevant VCE/NVENC/QuickSync 4K/60 "internet" option?

Thanks for the help here.

Please go to the form linked here and in my signature as it has the files and explicit instructions for render settings.

I just uploaded my results after coming across this benchmark thread. My RTX 3090 (assuming it is the GPU) has been blacking out all my monitors randomly for a couple seconds a couple times a day lately. No driver crash reporting or anything and everything seems to run well after it recovers. So I think the GPU is just getting bad. With 26 total renders of the whole project in various ways the only issue I encountered was one VP20 hang during a VCE render. All my NVENC renders were fine. I watched over each render and never noticed any glitches. From the looks of the results so far it looks like CPU is king so far for renders. That 4090 is sure looking nice though...

@hambonio Thanks for that. It appears your CPU has an iGPU (Radeon Graphics) which is how you did a VCE render, so I changed no to yes for integrated graphics.

Is it right that you had dynamic ram preview at 3232 for 2 runs and 2121 for NVENC?

If you disable legacy AVC do you get any better performance? As is you are almost at the top of the list which is expected for hardware as good as yours.

Hmm, odd about the blackout glitches. Perhaps try some benchmarks/stress tests on the GPU and update to the latest studio driver (released yesterday) and see if that helps? Your power supply has plenty of headroom for the 3090, right? It can have very high transient spikes from what I've read.

For VCE vs NVENC, the latter has some kind of bottleneck in VEGAS that results in modest speed reductions. If you use NVENC through Voukoder it loads the GPU more consistently. Still all your times are excellent so I think you can use whatever one you prefer.

If you feel like doing a second benchmark there's one linked in my signature that works in VP 16+. It is also GPU intensive but with different Fx. I'd just do a AVC UHD render and my guess is you'd be at the top of the list. I'm curious if your system could play that one back at full framerate (essentially breaking the benchmark as once you're at 25fps systems can no longer be ranked).

@hambonio Thanks for that. It appears your CPU has an iGPU (Radeon Graphics) which is how you did a VCE render, so I changed no to yes for integrated graphics

If VCE requires the integrated graphics then I guess I used it. I have it turned off everywhere else. I never do VCE renders anyhow.

Is it right that you had dynamic ram preview at 3232 for 2 runs and 2121 for NVENC?

It should be 3232, sorry typo and was a late night. I have it set at 5% from 64GB RAM.

If you disable legacy AVC do you get any better performance? As is you are almost at the top of the list which is expected for hardware as good as yours.

As far as I know, Legacy AVC really only affects timeline playback right? I tried with NVIDIA enabled for renders and they were pretty much identical as using Legacy AVC and I know my timelines play much smoother with that option enabled.

Hmm, odd about the blackout glitches. Perhaps try some benchmarks/stress tests on the GPU and update to the latest studio driver (released yesterday) and see if that helps? Your power supply has plenty of headroom for the 3090, right? It can have very high transient spikes from what I've read.

I use a 1200W power supply and the current power supply is less than 6 months old. I do have warranty through Microcenter on the RTX 3090 until November so I will just keep a watch on it. It is only a minor annoyance at the moment. If it does get bad enough I can pretty much almost do an even trade for a 4090.

I also have a PC with a Ryzen 9 3900x and an RTX 3070. That one I am nearly positive is giving me transient spikes as a couple times a day for the past 6 months just black screens and restarts on me. That is also a 1200W PSU - and the 2nd PSU in that build with no help. Oh also 2nd motherboard (It was a Crosshair VIII Hero, and now a TUF). So it seems I am troubleshooting a lot of things and not finding the root of the problem. Silver lining is I put a cheap CPU and RAM in the Crosshair VIII Hero now, and it is working great as a NAS with a ton of SATA ports available inside and 2.5GB ethernet.

I have a warranty on the RTX 3070 until next March. I am just trying to get as much life out of these as possible before I am annoyed enough to return the GPUs. Oddly, the restarts never happen when the PC is being moderately/heavily used. It almost always happens when there is light use like just web browsing.

For the record I don't have exact timings but I did try this project on that PC last night before I headed to bed and the Mainconcept was ~2:30, and nVENC was ~2:00. I never was able to write them down or try multiple renders as it is a shared PC and my wife was using it at the time. But if I get a chance I will run all these on that PC as well.

For VCE vs NVENC, the latter has some kind of bottleneck in VEGAS that results in modest speed reductions. If you use NVENC through Voukoder it loads the GPU more consistently. Still all your times are excellent so I think you can use whatever one you prefer.

Voukoder is pretty much all I use anymore. I think I had a bit faster render (and higher quality visually) when I tried it last night but also did not write it down. I figure that is not in the test project as it is not native to Vegas.

If I also remember my first NVENC render with MagixAVC was 1:55 but after that they were much faster. So I am also curious to see whether I can get a better time with the 3070 with multiple runs.

If you feel like doing a second benchmark there's one linked in my signature that works in VP 16+. It is also GPU intensive but with different Fx. I'd just do a AVC UHD render and my guess is you'd be at the top of the list. I'm curious if your system could play that one back at full framerate (essentially breaking the benchmark as once you're at 25fps systems can no longer be ranked).

I will have a look at this project as well.

I uploaded a couple of results from the VP16 project. Also watching hwinfo with all these renders today I realized how bad mainconcept is. With VCE and NVENC the CPU stays below 110W and under 70C. With Mainconcept I was hitting 235W and CPU was hitting my target max of 90C almost the entire time. Not efficient at all! Also I loaded the VP20 project again and tried some Voukoder renders. H264 NVENC and HEVC NVENC. CQ of 15 then CQ of 23 for each. Each of those renders landed in the 1:10-1:15 range. I guess the HEVC render surprised me the most. I don't do a lot of that within vegas, but use HEVC through Handbrake to compress all my raw footage when I am done with a project for archiving, and it is not nearly as fast. Playback of the rendered files were at full framerate.

Thanks, updated the dynamic ram preview amount.

Legacy AVC affects timeline playback and renders as before it can encode it has to decode the media. Whether it's better to have your Radeon graphics do it or send it to one of your many CPU cores may depend on the media in use; worth testing both ways.

It's not surprising that a CPU only render will tax the CPU and I don't think that's a problem per se. I use Mainconcept or Voukoder x264 for final renders given the increased size efficiency over GPU renders (better quality for the same bitrate) and definitely hear the CPU fan kick in. I have cooling and voltage and clocks set to handle it so I don't have stability issues with long renders.

Your times are respectable here. One value I found for this benchmark is that when VEGAS releases a new build and new settings or I change drivers, etc. I see if there are any changes to my times. Sometimes it can be useful as a diagnostic when something feels wrong.

I should mention that for a while I experimented with dynamic ram preview when I built this system. I WAS getting a lot of renders with weird glitches if I increased it too much, especially in the top 15%-20% of the screen. So I just set it to 5% now and that seems to play back just as smoothly as anything I had set 10%-50% of my RAM. Maybe that has to do with non QVL RAM, I don't know. But it was messing with my renders. I don't have any issues otherwise and any RAM tests I have done have all passed, and I have not seem to have any issue at 5% DRP.

I've seen similar things and will set DRP higher temporarily if I'm making extended previews. Its utility as a preview buffer is limited and the potential to conflict with renders real so I generally just have it at the default 5% in VP 19+. I don't think there's any flaw with your ram it's an issue within VEGAS.

Thanks. I did try temporarily setting to 25% yesterday and was getting glitches all over the preview window. I set back to 5% and they were still there. I restarted VP20 and still there. The only way I was able to completely drain that from the RAM it seems was to restart my PC and now working fine again at 5%. Good and bad to know its not likely my system, but VP.

I was just trying to see if anything had improved going that route since I built in Oct.

Hi - I just wanted to say that I'm immensely grateful to each one of you who have taken the time to dissect and scrutinize my project here in this thread. Seeing the extensive performance tests, discussions, and praise has humbled me beyond words. Despite being nowhere near as technically adept as you, I'm thrilled that all of you in this community is going all-in with testing and providing potentially valuable data for the developers.

There's something I'd like to highlight, particularly for those new to VEGAS Pro. You might find the discussions here about CPU/GPU processing, decoding specifications, QSV/NV/MC etc. quite complex, but it's not a prerequisite for success here. All you need to create in VEGAS Pro is a good idea, passion, and a goal. All my projects on this forum, including this one, were created on a midrange 2017 Acer laptop with no GPU acceleration, no external plugins or anything else fancy. It wasn't a walk in the park, but I hope it illustrates that you don't necessarily need advanced technical knowledge or a high-end computer to do cool stuff in VEGAS Pro.

Again, a huge thank you to all contributors to this thread. I've been following everything and thoroughly enjoyed your experiences with the project, especially how the AI-powered features and other demanding plugins fare on various setups. Thanks, everyone!

A huge shout-out to @RogerS who's been keeping track of the benchmarking results through his data collecting form. Makes it way easier to compare the results. Also @Howard-Vigorita who seriously took his time to manually transfer this project from VEGAS Pro 20 to VEGAS Pro 19. Wow...
And of course @fr0sty who started this thread in the first place and made it all happen.

Hi

On Saturday I played around with the render settings and reported one of them which was 3:15.

I'm running Vegas Pro 20 Build 411.

I thought I'd have another go today but I'm getting results in the 9+ minutes.

I have a Ryzen 5950X and have checked all the settings including GPU acceleration being on. I have tried with max 75% dynamic RAM, 0% and 5% with no real change. This is after a reboot with only Firefox and Task Manager open. I have even set Priority to High with no change. The GPU bounces around between 0% and 50% and the CP{U never gets higher than 7% at it's base rate of 3.4GHz.

There were two windows updates updates I installed this morning; KB5028166 and KB5028937.

Is there anything I've missed

Hi @Mav6215. 3:15 is a reasonable time. 9 minutes not so much. For dynamic ram preview put it back on the default as too high amounts can limit the ability of VEGAS to access ram for everything else and too low can dramatically worsen render times. In the benchmark the amount was listed as 32MB which is lower than recommended.

For priority normal is the appropriate level.

Go to help/ driver update in VEGAS and check you have the latest drivers. NVIDIA put out a new studio driver recently.

I'd disable legacy AVC decoding in preferences, file io. Is the reason you have that checked stability issues?

GPU should bounce around given the nature of the project. CPU use- 7% is low. You are sure there aren't other background processes, etc. taking resources away? Try another computer restart just to doublecheck.

You can try a tool like HWInfo64 to watch your CPU while it's being used and make sure it's not hitting power or thermal limits. Change the render from NVENC to Mainconcept and the CPU should be close to 100% the whole time.