Page 1 of 1

Extremely slow render of simple sequence

PostPosted: Wed Jun 13, 2018 6:23 am
by Stephen Dixon
I have a basic fusion comp that I've done from a timeline sequence. It involves one transform node, a colorgain node, a still image stretched out over the length of the sequence and a couple of merges. All the footage and the final result is 1080p.

I was basically giving fusion a try on something super simple, but it seems to have failed dismally: rendering is so slow as to be almost unusable - I'm getting a frame rate of around 2 fps. Normal rendering from resolve on this machine when I do grades is around 50-70fps for 1080p.

Is this just because it's still in beta, is it always this glacial, or am I doing something wrong?

2018-06-13 16_13_13-DaVinci Resolve by Blackmagic Design - 15.0.png
the nodes
2018-06-13 16_13_13-DaVinci Resolve by Blackmagic Design - 15.0.png (66.9 KiB) Viewed 1774 times

Re: Extremely slow render of simple sequence

PostPosted: Thu Jun 14, 2018 2:07 am
by Peter Chamberlain
System info will help.. likely resources .. and it will require a render.. often slow.
viewtopic.php?f=32&t=71894

Re: Extremely slow render of simple sequence

PostPosted: Thu Jun 14, 2018 5:06 am
by Stephen Dixon
Dual Xeon E5-2690 20 core @ 3.0 Ghz (40 physical processors)
64 Gb Ram
2×NVIDIA 1080 GTX Titan Black 6Gb
24 Tb hardware RAID5 media storage, 1Tb SSD system drive, crystal mark drive tests below:
Windows 10 64 bit. Resolve 15 b4

Sequential Read / Write (Q= 32,T= 1): 503.618 MB/s / 329.884 MB/s
Random Read / Write 4KiB (Q= 8,T= 8): 14.876 MB/s / 2.562 MB/s
Random Read / Write 4KiB (Q= 32,T= 1): 12.260 MB/s / Write 2.492 MB/s
Random Read / Write 4KiB (Q= 1,T= 1): 1.956 MB/s / 2.067 MB/s

The footage is a couple of 10-bit png sequences, 1920×1200, stored on the media RAID, and rendered back to it.

If this machine doesn't have the resources to get a good render speed, then nothing does.

Re: Extremely slow render of simple sequence

PostPosted: Thu Jun 14, 2018 5:36 am
by Peter Chamberlain
Agree, your system seems ok. Can u export the comp and PM to me so I can check if it’s something we can improve now or if it’s something we need to write faster algorithms for