I am using latest nvidia geforce driver 431.60.
Didn't try Nvidia studio driver. I could check that if that will help out.

Nvidia says that driver is specifically optimized for Vegas.

So only solution is to turn it off, but then I lose a performance.

Your problem is with rendering, what loss of performance do you have when ending that accelerationtask just before rendering? Did you try that already?

No mention of that Hardware I/O setting for NVIDIA? Would that help? Has helped me.

No mention of that Hardware I/O setting for NVIDIA? Would that help? Has helped me.

That's meant for decoding, not encoding.
AFAIK rendering is encoding.

It would still affect the render process because rendering consists of decoding a frame, processing the frame (size, color, fx, frame blending, etc.) and encoding the frame.

Maybe Marco, but I did only see advantages with preview of heavy files. My rendertimes did not change ( that was already very quick with NVENC 😃)

Maybe Marco, but I did only see advantages with preview of heavy files. My rendertimes did not change ( that was already very quick with NVENC 😃)

@j-v - Do you have any numbers that support your qualitative observation? As @Marco. is saying that I/O is also about encoding, well it would, wouldn’t it. And again, you’ve not said why NOT having the I/O would be an advantage or disadvantage? I’m suggesting, or rather asking, the opposite, having that engaged adds to better workflow. And possibly only @VEGASDerek could jump in here and conclusively agree or deny. 😉

Maybe Marco, but I did only see advantages with preview of heavy files. My rendertimes did not change ( that was already very quick with NVENC 😃)

Correction, that behaviour I saw on my laptop.
But now I'm back from my long vacation in Norway and I tested it now on my desktop.
Rendertime there of a 1 minute render to FHD Magix AVC (NVENC) is faster in VP17 than in VP16 with exactly the same template.
1 minute in VP16 renders in 54 sec.
The same in VP 17 with my already changed and showed File I/O setting renders in 44 sec., almost a gain of 20%!

I'm starting to think the problem could be MagicYUV codec that I use for recording game videos. I really don't see any performance gain with GPU enable / disabled, not even in a preview. As for rendering my CPU does the great job 30 sec of 4k video renders in 1 min (Magix HEVC codec), what i think is totally ok.
 

GPUs are only capable of hardware decoding AVC and HEVC formats, so MagicYUV isn't a format it is capable of decoding, though I do know it plays nice with Vegas, as it is one of the codecs that comes bundled with Happy Otter Scripts. It's a high quality intermediate codec.

Rendertime there of a 1 minute render to FHD Magix AVC (NVENC) is faster in VP17 than in VP16 with exactly the same template.
1 minute in VP16 renders in 54 sec.
The same in VP 17 with my already changed and showed File I/O setting renders in 44 sec., almost a gain of 20%!

@j-v - Happy Days! Nothing like Quantitative Evidenced-Based Results. Now, @j-v, having your GPU taking some, or a greater part of the heavy lifting AWAY from the CPU, this allows the CPU to get on with the “back-office” Job. And again KUDOS to MAGIX Engineers to jump in with both feet and wrangle the Hardware<>Software conundrum. I’m wagging ALL my tails!

Last night I installed Particle Illusion.....stunning frame rates.

@j-v - Happy Days!

😁😃😄👍
and 👌 for the Vegas engineers!

@j-v @Grazie Happy days indeed. I had assumed that it wasn’t working on my systems because there was no change in render times on the J-V SampleProject 4K? Maybe the clips are not fully compliant? and of course theres other stuff in the project as well.

Anyway I did a quick test on laptop and PC using 19s clip x 3. The laptop showed the biggest benefit, 43%, PC 21%, all when rendering UHD h264 clips to FHD h264, really excellent.

There's another Vegas acceleration aspect I've been meaning to explore more deeply but have not gotten to benchmarking yet... plugin acceleration. Plugins are available in 2 forms: gpu accelerated and 32-bit floating point. I believe based on my debugging issues in the past related to gpu-plugins causing snow in 60i rendering for dvd and bluray, that selecting a gpu accelerated plugin then turning off the gpu in video settings causes the plugin into operate as if it was cpu-only. But I'm not clear on exactly what happens in Vegas with 32-bit floating point plugins on different platforms like the i7, i9, and e5 which all handle floating point differently... i7 uses avx2, i9 uses higher performance avx512, and e5 Xeon uses its legacy x87 fpu. Anybody got any inside info on this?

I have i9 7980xe. Also I have try out 32bit floating point (full range)
I see huge diff if I am using compositing gamma 1.000 (linear) vs 2.222 (video)

Linear is working like that I have celeron CPU in my machine (it almost need like 40 min), while video is rendering 4k 60fps 30sec video in 1 min.

@j-v @Grazie I did a small test on PC also using Intel instead of Nvidia in the file I/O tab.

Result was that in my case the Nvidia was the best option for render times. The render times turned in using Intel were the same as VP16 and VP17, as if no benefit at all from Intel QSV HW Decoding.

HW Acceleration was left in both cases as Nvidia.

@Former user Have you got the % figures?

I've created a new post for this Intel QSV non decoding issue … https://www.vegascreativesoftware.info/us/forum/vp-17-intel-qsv-hd-630-igpu-not-used-for-decoding--116762/

@Grazie Yes it was a 0% render time change using Intel QSV decoding in VP17 vs simply using VP16. I suspect that the "Intel Graphics 630" is not being used for decoding. Note that unchecking QSV in "General" simply makes it unavailable in the render templates for testing.

I have confirmed that the Intel QSV HW encoding is working using the Red Car test with a gain of 5s and this 57s test with a gain of 3s.

I can also confirm that while observing the Intel QSV gpu in Task Manager it never shows any decoding activity.

SECTION 1

19s duration clip length x 3 = test project duration of 57s, UHD 25fps.

HW Acc. = Nvidia for all.

VP17 .. UHD to FHD using Nvidia in file I/O, render time = 15s ..... 21% faster render.

VP17 .. UHD to FHD using Intel QSV in file I/O, render time = 19s

VP17 .. UHD to FHD using no file I/O decoding, render time = 19s

VP16 .. UHD to FHD …………………………………………. render time = 19s

SECTION 2

I tried something else with respect to the non-decoding of Intel QSV on my system. I set HW Acc to Intel QSV instead of Nvidia. Maybe decoding needs HW ACC. to have Intel selected?

I then rendered out the same test project with Intel QSV selected in the File I/O tab, and tested with HW Decoding on or off. The result is no change, HW decoding using Intel QSV is not working on my PC, but QSV HW encoding is.

19s duration clip length x 3 = test project duration of 57s, UHD 25fps.

HW Acc. = Intel QSV

File I/O =  Intel QSV

VP17 .. UHD to FHD, "Enable Hardware Decoding" in file I/O = ON, render time = 47s.

VP17 .. UHD to FHD, "Enable Hardware Decoding" in file I/O = OFF render time = 47s

@Former user Have you got the % figures?

Mine are for 1 minute render 2,7K to my usual AVC FDH 50p with NVENC with some cropping in Pan/Crop
VPro 16 .................................................62 sec (100%)
VPro 17 File I/O set to QSV (Intel) .........57 sec (8% less)
VPro 17 File I/O set to NVDEC (Nvidia)..43 sec (30% less)

Stunning guys!

@j-v @Grazie I've added in the % improvement in my previous post, yes Grazie, excellent, and no doubt improved timeline performance, just no way to properly test it.