Page 1 of 1

Didn't know there were preferences test says faster with CPU only?

Posted: Fri Dec 23, 2022 11:55 pm
by dfphoto
Hi I own Neat video and was surprised to find out the test function says 3x faster using the CPU only vs the GPU or both.
This is on my older mbp with the AMD560 4gb.... I've been struggling the last 2 weeks did a wedding video for a corporate client 3 camera 10 bit 4.2.2 in 4k...

The video I broke up into segments and exported them out in FCP to 4k from 4k and then after final assembly I reduce the size to 1080p prores... and h.264

Since this was an event it was very dark and I needed to use Neat Video quite a bit 5.0 on my MBP M1 max 32gb ram and the export is killing me as it is taking 9-15 hours per with about 1 hour of neat video on various clips or segments.

If the M1 max is similar to my test above the stock preference in Neat is for gpu can anyone verify if CPU only is the fastest way to use Neat Video?

Re: Didn't know there were preferences test says faster with CPU only?

Posted: Fri Dec 23, 2022 11:57 pm
by dfphoto
BTW I create a custom profile and select best quality - slowest and I reduce the amount of denoise to about 30% if that matters....

Re: Didn't know there were preferences test says faster with CPU only?

Posted: Sat Dec 24, 2022 1:43 am
by dfphoto
Okay got a chance to test on my M1 and in 4k 32bit the max is 2.44 frames per sec and that is with both cpu/gpu with cpu only it was .98 frames 2.3 frames on gpu only and 2.44 with both... disappointing I guess that's why it took 9-15 hours export. Looks nice, but man it killed me to export this out... 2.2 hours video 9-15 hours export in compressor and that's with a 16tb SSD RAID 0 drive...

Re: Didn't know there were preferences test says faster with CPU only?

Posted: Sat Dec 24, 2022 11:11 am
by NVTeam
1. Please change the quality settings back to Normal, this will improve the speed significantly.

2. If your final export needs to be 1080p then it is possible to speed up rendering by changing the frame size to 1080p before heavy processing. Please see this article for details.

Thank you,
Vlad