Some Vegas Pro 15 vs 14 render times

RealityStudio wrote on 8/28/2017, 5:55 PM

Just bought and installed Vegas Pro 15. My system is pretty basic:

AMD Ryzen 1700

NVidia 980ti

16gb ram

Windows 10

Latest drivers for everything, and I've left everything stock, no overclocking or tweaking, I've found that just causes unnecessary headaches. Likewise with Vegas 15 I've tweaked nothing, left it all at stock install. Made a test clip that is 1:12 long and it has a watermark and Film Convert color grade on it. Source footage is 100mbs 4k from my Sony AX100 camera. Encode results:

Vegas 14, 4k, Main Concept: 7:14, 236mb file

Vegas 14, 1080p, Main Contept, 2:24, 116mb file

Vegas 15, 4k, Magix AVC, 6:56, 244mb file

Vegas 15, 4k, Magix AVC NVENC, 2:08, 212mb file

Vegas 15, 1080p, Magix AVC NVENC, 0:51, 109mb file

So for me on 4k footage I went from 7:14 on Vegas 14 to 2:08 on Vegas 15 with gpu encode. Everything is stock, I tweaked nothing on both Vegas or my PC, everything seems to run solid. I've very happy, no issues so far. Footage for what I film at least looks good quality with NVENC so I'm sticking with it.

Comments

Wolfgang S. wrote on 8/28/2017, 6:34 PM

So for me on 4k footage I went from 7:14 on Vegas 14 to 2:08 on Vegas 15 with gpu encode. Everything is stock, I tweaked nothing on both Vegas or my PC, everything seems to run solid. I've very happy, no issues so far. Footage for what I film at least looks good quality with NVENC so I'm sticking with it.

I have very similar results here, but with a much weaker Quadro K4200: -55% down in render time. I think that is great!

RealityStudio wrote on 8/28/2017, 7:53 PM

I'm wondering how much faster a 1080i would be than my 980ti. According to the wiki page on NVENC:

" It also doubles the encoding performance of 4K H.264 & HEVC when compared to previous generation NVENC."

That's talking about a Pascal based 1080ti compared to my Maxwell based 980ti. Tempting to try one.

fr0sty wrote on 8/28/2017, 8:42 PM

I ran some benchmarks using the Vegas 11 benchmark project that was converted to 4K

Ryzen 7 1800x (stock speed)

32GB 3000mhz DDR4 (stock speed)

Nvidia GTX 970 GPU

Results:

4K 29.97fps Magix AVC NVENC: 6:04

4K 29.97fpsMagix AVC CPU: 8:11 (Vegas 14 MC AVC Speed: 8:17)

Magix 29.97fps AVC 1080p NVENC: 2:45

Magix 29.97fps AVC 1080p CPU: 3:17 - (Vegas 14 MC AVC 1080: 3:14)

XAVC-I 3840x2160 29.97p: 20:11 - (Vegas 14 Speed: 21:52)

XDCAM EX 1080p 29.97fps 2:36 (Vegas 14 Speed: 2:28)

DVD Architech DVD 480i: 2:45 (Vegas 14 Speed: 2:27)

Peter_P wrote on 8/29/2017, 2:13 AM

I ran some benchmarks using the Vegas 11 benchmark project that was converted to 4K

Ryzen 7 1800x (stock speed)

32GB 3000mhz DDR4 (stock speed)

Nvidia GTX 970 GPU

Results:

4K 29.97fps Magix AVC NVENC: 6:04

I just rendered the same Vp11 benchmark upscaled as a 3840x2160 29.97fps project with the new Magix AVC/AAC encoder :

On my i7-6700k 4.5GHz 16GB RAM AMD R7 250 it takes only 1:28 (F=1.35) to render and the UHDp30 output file is 172MB in size.

fr0sty wrote on 8/29/2017, 2:45 AM

Did you first upscale the assets to 4K? I first turned the 3 video files into 4K MXF files (each weighs in at about 7-8gb after conversion) and then opened the PNG files in photoshop and also upscaled them to 4K.

If so, that's a pretty big jump in render speed by using Intel QSV.

Peter_P wrote on 8/29/2017, 2:56 AM

Did you first upscale the assets to 4K?

No, I used the original footage and just changed the project setting to UHDp30. So rendering includes the upscaling.

Yes, I would really miss the QSV speed (specially on HEVC render, that I mostly use for UHDp30) and the combination I currently us, also allows to preview UHDp50 from a GH5 with full frame rate.

RealityStudio wrote on 8/29/2017, 10:56 AM

Would be cool to get QSV vs NVENC encode test using the exact same 4k source files and project. Alas I don't have any Intel setup to test on.

fr0sty wrote on 8/29/2017, 1:41 PM

Did you first upscale the assets to 4K?

No, I used the original footage and just changed the project setting to UHDp30. So rendering includes the upscaling.

Yes, I would really miss the QSV speed (specially on HEVC render, that I mostly use for UHDp30) and the combination I currently us, also allows to preview UHDp50 from a GH5 with full frame rate.

That's why mine took so much longer. I was working with much bigger and higher quality files on the timeline.

Peter_P wrote on 8/30/2017, 10:23 AM

Did you first upscale the assets to 4K?

No, I used the original footage and just changed the project setting to UHDp30. So rendering includes the upscaling.

Yes, I would really miss the QSV speed (specially on HEVC render, that I mostly use for UHDp30) and the combination I currently us, also allows to preview UHDp50 from a GH5 with full frame rate.

That's why mine took so much longer. I was working with much bigger and higher quality files on the timeline.

You can easily check this out by using the same setup I used and include the upscaling to the rendering. Would be interesting if this really is faster.

fr0sty wrote on 8/30/2017, 1:10 PM

Unfortunately I no longer have the original non-upscaled assets.

Peter_P wrote on 8/31/2017, 1:13 AM

Probably you or someone can find the original Vp11_benchmark in google cash or on some other location for download.

NickHope wrote on 8/31/2017, 1:23 AM

https://web.archive.org/web/20131113061953/http://download.sonycreativesoftware.com/whitepapers/vp11_benchmark.zip

Peter_P wrote on 8/31/2017, 1:55 AM

Thanks Nick, coud you please pin this in your very helpfull FAQs ?

NickHope wrote on 8/31/2017, 3:36 AM

Had already updated the link in section 4 of the GPU FAQ post, but I've just added it to the main FAQ post, currently #29.

Peter_P wrote on 8/31/2017, 4:04 AM

Thanks.

rkm wrote on 9/9/2017, 5:51 AM

Hi all, the linked archive.org .zip file doesn't open (archive errror), tried downloading twice... and the official download.sonycreativesoftware. links you floating in other forum threads are offline. Any chance someone can re-upload this benchmark file to a file service and post the link? Would be highly appreciated :-)

Thanks

Rob

Former user wrote on 9/9/2017, 11:19 AM

I updated my tests with the un-modified Red car test. Theres a .txt file in the dropbox also with all of the test results.

The bottom line, for my system, others may have different results, is Nvidia gtx 1080 for timeline preview acceleration and Nvenc for HW render.

https://www.vegascreativesoftware.info/us/forum/someone-with-threadripper-1950x-or-skylake-x-7900x-please-test-this--108381/#ca668333