Ben_Nicholls
11th September 2017, 10:11
Is it just me or has anybody else noticed that when trying to run MCTemporalDenoise on Avisynth+ it seems very slow? I am barely getting 1fps processing 720p24 video using Avisynth+ 2508 x86.
It is worth noting that (with enough patience) the output video looks correctly filtered.
Initially I tried the MCTD mod v4.3 by AmjadSONY, but given the speed issues I reverted to LaTo's last official MCTemporalDenoise v1.4.20. Didn't seem to change anything.
My machine isn't a supercomputer, but it is still decent. I am using a:
Intel Core i7 4820K clocked to 4.6GHz with Hyper-Threading enabled (water cooled with Corsair H100i)
ASUS x79 Deluxe
32GB 2133MHz Corsair DDR3 RAM
Samsung M.2 950 Pro 500GB NVMe (boot drive)
2x Samsung EVO 1TB (RAID 0, used largely for video and audio editing)
2x Seagatte Barracuda 3TB (RAID 1, for long term storage)
ASUS GeForce GTX 780 Ti OC
Windows 10 Professional x64
I have checked my Avisynth+ settings and it (and its dependent plug-ins) are all set to use MT. When I load the script into VirtualDub v1.10.4 (x86) and do a preview pass all 8 cores are in use. With Avisynth v2.6.0.5 (SEt's MT version) it was never very fast (maybe something in the region of 5-10FPS for 720p24). However since then the system has actually been upgraded somewhat (it it is in need of moving to one of the newer platforms... especially now a lot lf the plugins have an AVX2 build and my 4820K only handles AVX1. am going to wait and see how X299 Socket 2066 turns out).
Anyway, the videos are stored on the Samsung EVO SSD RAID 0 (which, while not quite up to the M.2's standard are still very fast, something around 950MB/s read 850MB/s write).
Any ideas what could be causing MCTD to perform so slowly in comparison to AviSynth v2.6.0.5 MT? I have always used MCTD heavily for difficult sources. SMDegrain's results are pretty good and it runs much faster. I suppose I may have ti, but I found MCTD to be by far the best de-nosier out there (the post processing optipns are also very convenient).
Any advice would be great! :)
It is worth noting that (with enough patience) the output video looks correctly filtered.
Initially I tried the MCTD mod v4.3 by AmjadSONY, but given the speed issues I reverted to LaTo's last official MCTemporalDenoise v1.4.20. Didn't seem to change anything.
My machine isn't a supercomputer, but it is still decent. I am using a:
Intel Core i7 4820K clocked to 4.6GHz with Hyper-Threading enabled (water cooled with Corsair H100i)
ASUS x79 Deluxe
32GB 2133MHz Corsair DDR3 RAM
Samsung M.2 950 Pro 500GB NVMe (boot drive)
2x Samsung EVO 1TB (RAID 0, used largely for video and audio editing)
2x Seagatte Barracuda 3TB (RAID 1, for long term storage)
ASUS GeForce GTX 780 Ti OC
Windows 10 Professional x64
I have checked my Avisynth+ settings and it (and its dependent plug-ins) are all set to use MT. When I load the script into VirtualDub v1.10.4 (x86) and do a preview pass all 8 cores are in use. With Avisynth v2.6.0.5 (SEt's MT version) it was never very fast (maybe something in the region of 5-10FPS for 720p24). However since then the system has actually been upgraded somewhat (it it is in need of moving to one of the newer platforms... especially now a lot lf the plugins have an AVX2 build and my 4820K only handles AVX1. am going to wait and see how X299 Socket 2066 turns out).
Anyway, the videos are stored on the Samsung EVO SSD RAID 0 (which, while not quite up to the M.2's standard are still very fast, something around 950MB/s read 850MB/s write).
Any ideas what could be causing MCTD to perform so slowly in comparison to AviSynth v2.6.0.5 MT? I have always used MCTD heavily for difficult sources. SMDegrain's results are pretty good and it runs much faster. I suppose I may have ti, but I found MCTD to be by far the best de-nosier out there (the post processing optipns are also very convenient).
Any advice would be great! :)