Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > High Efficiency Video Coding (HEVC)

Reply
 
Thread Tools Search this Thread Display Modes
Old 3rd May 2019, 10:48   #21  |  Link
excellentswordfight
Lost my old account :(
 
Join Date: Jul 2017
Posts: 322
Quote:
Originally Posted by TEB View Post
@sneaker

25p x 4 sec gop = 100

Middle profile of a long stack/ladder

bench : 14.4x Realttime
Well I think he understood that part, but why is the min-keyint higher then keyint. And why use level 5.1?

Anyway, did you try lowering ctu and merange?

edit:

"ffmpeg.exe" -i "1080p25_input" -vf "scale=960:540" -f yuv4mpegpipe -strict -1 - | "x265.exe" --y4m --input-depth 8 --output-depth 10--preset slow --profile main10 --level-idc 31 --crf 23 --ctu 32 --merange 26 --keyint 100 --min-keyint 25 --rc-lookahead 100 --vbv-maxrate 3000 --vbv-bufsize 3000 --no-open-gop --colorprim bt709 --transfer bt709 --colormatrix bt709 --range limited - -o NUL

I guess this is some what simlar to the video settings you are going for, I get 60% CPU usage on 24T with those.

edit2.
Wait what, 14x realtime!? Are you hitting 350fps? I'm hitting close 2x realtime with the settings above, ~45fps.

Last edited by excellentswordfight; 3rd May 2019 at 14:04.
excellentswordfight is offline   Reply With Quote
Old 3rd May 2019, 11:28   #22  |  Link
~ VEGETA ~
The cult of personality
 
~ VEGETA ~'s Avatar
 
Join Date: May 2013
Location: Planet Vegeta
Posts: 155
Quote:
Originally Posted by RanmaCanada View Post
Not worth the money. I recently moved from an E5-2670 to a Ryzen 2700 and I literally almost doubled my frame rate while cutting my power usage by a good third (150 watts while encoding vs 230 watts). If I overclock the 2700 to 2700x speeds, I do double my performance from my old Xeon at the same power usage. The sheer lack of AVX 2.0 makes them horrible for x265. There is a thread by Sagitarre? that properly benches processors for x265. If you check the thread you'll see that all AVX processors are beaten by similarly spec'd AVX 2.0 processors. Even my old 8 core was beaten easily by a quad core with AVX 2.0.

I'd strongly suggest you wait till Zen2 comes out before you make your purchase. You will either be able to get a new system that will destroy the dual Xeon and sip power, or someone's old stuff because they wanted to upgrade. Should be anywhere from 1-2 months at this point.
It is about 220$ only which is half the price of 2 of those xeon cpus, are you sure it gives the same processing power for x264 and x265? since I encode mostly in x264 10-bit and it is what is important to me the most.

Also, I liked the xeon build because it allows me to run engineering software (solidworks, rendering stuff,etc...) better than normal cpu. I also do some light gaming (league of legends) so although xeons are not suitable for gaming, they are enough for that if I pair them with a good gpu (say 1080ti gtx or something). that was the plan and it won't be applied before 6 months or so.

So is this ryzen 2700 any good for gaming and big demanding software? how much fps do you get while encoding (just a general estimate) 1080p material? rough figures are welcome.

edit: kindly give me the link to that benchmarking thread.
~ VEGETA ~ is offline   Reply With Quote
Old 3rd May 2019, 12:14   #23  |  Link
excellentswordfight
Lost my old account :(
 
Join Date: Jul 2017
Posts: 322
Quote:
Originally Posted by ~ VEGETA ~ View Post
Also, I liked the xeon build because it allows me to run engineering software (solidworks, rendering stuff,etc...) better than normal cpu.
There is no magic on the xeon line for that stuff, its basically the ECC support (which is not cricital for those workloads) and the fact that in past you almost had to use them to get the core count up. But in reality you could actually lose performance because they lacked in singel thread performance compared to the consumer line, cause most workstations are not used for rendering 24/7 so interactive workloads are faster on consumer chips a lot of the time. This have been somewhat fixed with the new Xeon-W line though.

Quote:
So is this ryzen 2700 any good for gaming and big demanding software? how much fps do you get while encoding (just a general estimate) 1080p material? rough figures are welcome.

edit: kindly give me the link to that benchmarking thread.
The most promising part of the upcomming Ryzen3 is that they should no longer have a big AVX disadvantage compared to intel (9700k beats or is close to 2700X in x264 and x265 because of it for example). So they should give very nice value for both encoding and general workloads like gaming and professional software suits cause you get an high thread count (for an consumer line) and decent singelthread performance.

Last edited by excellentswordfight; 3rd May 2019 at 12:46.
excellentswordfight is offline   Reply With Quote
Old 3rd May 2019, 15:05   #24  |  Link
~ VEGETA ~
The cult of personality
 
~ VEGETA ~'s Avatar
 
Join Date: May 2013
Location: Planet Vegeta
Posts: 155
Quote:
Originally Posted by excellentswordfight View Post
There is no magic on the xeon line for that stuff, its basically the ECC support (which is not cricital for those workloads) and the fact that in past you almost had to use them to get the core count up. But in reality you could actually lose performance because they lacked in singel thread performance compared to the consumer line, cause most workstations are not used for rendering 24/7 so interactive workloads are faster on consumer chips a lot of the time. This have been somewhat fixed with the new Xeon-W line though.


The most promising part of the upcomming Ryzen3 is that they should no longer have a big AVX disadvantage compared to intel (9700k beats or is close to 2700X in x264 and x265 because of it for example). So they should give very nice value for both encoding and general workloads like gaming and professional software suits cause you get an high thread count (for an consumer line) and decent singelthread performance.

I am planning to get a nice desktop computer later on, which is to be used for everything including encoding, engineering software (solidworks, autocad plant 3d, etc...), little gaming (not too huge ones) and so on.

Some people recommended Ryzen 2700x to me as the best for all-purpose PC, coupled with good nvidia GPU (1060 or more).

In encoding, I use vapoursynth with stuff like bm3d, f3kdb, line stuff, etc.. and encode 1080p material mostly and in x264-10bit. I have online.net atom 2750 dedicated server which is no were near good for encoding but I get it done since I don't really care for speed on a dedicated server.

However, I really want some nice fps if I wanna do it on my desktop right? I'd say 10 fps is nice and +20 fps is good for such a script. ofc, encoding to lossless then directly to 1080p and 720p seems faster.
~ VEGETA ~ is offline   Reply With Quote
Old 3rd May 2019, 15:32   #25  |  Link
Forteen88
Herr
 
Join Date: Apr 2009
Location: North Europe
Posts: 556
Quote:
Originally Posted by ~ VEGETA ~ View Post
I...and encode 1080p material mostly and in x264-10bit.
x264-10bit encodes are not GPU-decodable att all, so I would not recommend doing those encodes. x265-10bit encodes are fully GPU-decodable on Nvidia GF 960 graphicscards or better.
Forteen88 is offline   Reply With Quote
Old 3rd May 2019, 19:51   #26  |  Link
~ VEGETA ~
The cult of personality
 
~ VEGETA ~'s Avatar
 
Join Date: May 2013
Location: Planet Vegeta
Posts: 155
Quote:
Originally Posted by Forteen88 View Post
x264-10bit encodes are not GPU-decodable att all, so I would not recommend doing those encodes. x265-10bit encodes are fully GPU-decodable on Nvidia GF 960 graphicscards or better.
I am not sure I understood what you meant.

I use x264-10bit so I need good cpu to do it, but also should be good enough for gaming and stuff. is ryzen 2700x the one?
~ VEGETA ~ is offline   Reply With Quote
Old 3rd May 2019, 20:04   #27  |  Link
Nico8583
Registered User
 
Join Date: Jan 2010
Location: France
Posts: 851
Perhaps you should wait Ryzen 3 ?
Nico8583 is offline   Reply With Quote
Old 3rd May 2019, 22:53   #28  |  Link
~ VEGETA ~
The cult of personality
 
~ VEGETA ~'s Avatar
 
Join Date: May 2013
Location: Planet Vegeta
Posts: 155
The price is king too xD.
~ VEGETA ~ is offline   Reply With Quote
Old 3rd May 2019, 22:58   #29  |  Link
TEB
Registered User
 
Join Date: Feb 2003
Location: Palmcoast of Norway
Posts: 363
Quote:
Originally Posted by excellentswordfight View Post
Well I think he understood that part, but why is the min-keyint higher then keyint. And why use level 5.1?
The first one is a trick to get forced IDR at the start and dynamic Iframes on scenecut. Only syntax ive found that actually works..

5.1 was just some leftover code from UHD.. I can lower it since its not needed..

Anyway, did you try lowering ctu and merange?
Nope, i have zero clue what those 2 mean.

edit:

"ffmpeg.exe" -i "1080p25_input" -vf "scale=960:540" -f yuv4mpegpipe -strict -1 - | "x265.exe" --y4m --input-depth 8 --output-depth 10--preset slow --profile main10 --level-idc 31 --crf 23 --ctu 32 --merange 26 --keyint 100 --min-keyint 25 --rc-lookahead 100 --vbv-maxrate 3000 --vbv-bufsize 3000 --no-open-gop --colorprim bt709 --transfer bt709 --colormatrix bt709 --range limited - -o NUL

I guess this is some what simlar to the video settings you are going for, I get 60% CPU usage on 24T with those.

Strange that im not getting that..

edit2.
Wait what, 14x realtime!? Are you hitting 350fps? I'm hitting close 2x realtime with the settings above, ~45fps.[/QUOTE]

Well thats what the benchmark result said. Prores file is on a NFS share with 10gig interface

Last edited by TEB; 3rd May 2019 at 23:00.
TEB is offline   Reply With Quote
Old 4th May 2019, 00:04   #30  |  Link
excellentswordfight
Lost my old account :(
 
Join Date: Jul 2017
Posts: 322
Quote:
Originally Posted by TEB View Post
The first one is a trick to get forced IDR at the start and dynamic Iframes on scenecut. Only syntax ive found that actually works..
So I guess you want the IDR frames to be fixed at 4s intervals and scenecut I frames not to be IDRs?


Quote:
Nope, i have zero clue what those 2 mean.
From the docs:

"Maximum CU size (width and height). The larger the maximum CU size, the more efficiently x265 can encode flat areas of the picture, giving large reductions in bitrate. However this comes at a loss of parallelism with fewer rows of CUs that can be encoded in parallel,"

The benefit of the defaut value of 64 isnt that big with lower resolutions, so it can be a very nice trade off performance wise to lower it if you are seing low thread utilization. And with a lower CTU merange can also be lowered.

Quote:
Well thats what the benchmark result said. Prores file is on a NFS share with 10gig interface
Sorry, I missed that it was an decoder benchmark, I thought that it was your encoder speed! And in that case, the bottleneck shoudlnt be on the decoder.
excellentswordfight is offline   Reply With Quote
Old 18th June 2019, 21:03   #31  |  Link
mparade
Registered User
 
Join Date: Nov 2013
Posts: 577
I cannot saturate the cores of my threadripper 1950X even by running 4 pcs of 2160p encodes in parallel. At first, my machine is running out memory (16GB).

I am using preset slow and 1 frame thread in my encodes. My threadripper behaves completely differently
as the one of blublub's.


Any advice would be appreciated.
mparade is offline   Reply With Quote
Old 21st June 2019, 11:18   #32  |  Link
Atak_Snajpera
RipBot264 author
 
Atak_Snajpera's Avatar
 
Join Date: May 2006
Location: Poland
Posts: 7,806
Quote:
Originally Posted by mparade View Post
I cannot saturate the cores of my threadripper 1950X even by running 4 pcs of 2160p encodes in parallel. At first, my machine is running out memory (16GB).

I am using preset slow and 1 frame thread in my encodes. My threadripper behaves completely differently
as the one of blublub's.


Any advice would be appreciated.
Running four x265 encoders with 2160p source is just insane! No wonder you are running out of memory. One instance should be enough for 32 threads. What GUI are you using? Show us cpu usage in Process Hacker for each process spawned by your application (ffmpeg.exe , x265.exe and so on).
Atak_Snajpera is offline   Reply With Quote
Old 21st June 2019, 12:13   #33  |  Link
Forteen88
Herr
 
Join Date: Apr 2009
Location: North Europe
Posts: 556
@Atak_Snajpera. He's probably doing that for better image quality on the encodes.
Using many threads in x265/x264 equals worse image quality.

If the encode had a lower resolution than 2160p, I'd suggest lowering --me-range and --ctu.

Last edited by Forteen88; 21st June 2019 at 14:01.
Forteen88 is offline   Reply With Quote
Old 21st June 2019, 19:29   #34  |  Link
Asmodian
Registered User
 
Join Date: Feb 2002
Location: San Jose, California
Posts: 4,406
I always use --frame-threads 1 myself but I only need two encodes for my 20 threads and I have 32GB of memory.

Two frame threads has very little impact of quality, it may be worth it. Either that or more RAM.
__________________
madVR options explained
Asmodian is offline   Reply With Quote
Old 21st June 2019, 20:18   #35  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,988
There are occasional issues with more than 1 frame thread, I highly suggest anyone using vbv to only use 1 frame thread

There are also still edge cases where using vbv will cause a few frames of explosive blocking, mostly with HDR. I'm actually working right now on isolating the issue.
Blue_MiSfit is offline   Reply With Quote
Old 23rd June 2019, 21:35   #36  |  Link
mparade
Registered User
 
Join Date: Nov 2013
Posts: 577
Thanks for the answers.
I solved this "memory" issue by:

- switched memory mode to local from distributed;
- switched off SMT in AMD Ryzen Master;
- installing 4x8Gbyte RAM instead of 1x16GB;
- using 8 threads per encode (4 encodes are being run in parallel to reach 100% processor utilization and ~60% level of memory);
- using 1 frame thread per encode with preset slower;
- this way my 1950X System is working stable while encoding time is quite acceptable until I my new PC arrives.
mparade is offline   Reply With Quote
Old 24th June 2019, 09:43   #37  |  Link
Atak_Snajpera
RipBot264 author
 
Atak_Snajpera's Avatar
 
Join Date: May 2006
Location: Poland
Posts: 7,806
Quote:
Originally Posted by mparade View Post
Thanks for the answers.
I solved this "memory" issue by:

- switched memory mode to local from distributed;
- switched off SMT in AMD Ryzen Master;
- installing 4x8Gbyte RAM instead of 1x16GB;
- using 8 threads per encode (4 encodes are being run in parallel to reach 100% processor utilization and ~60% level of memory);
- using 1 frame thread per encode with preset slower;
- this way my 1950X System is working stable while encoding time is quite acceptable until I my new PC arrives.
You are losing a lot of performance by disabling SMT! NEVER DO THAT ON AMD's CPUS!
Atak_Snajpera is offline   Reply With Quote
Old 24th June 2019, 17:13   #38  |  Link
mparade
Registered User
 
Join Date: Nov 2013
Posts: 577
Quote:
Originally Posted by Atak_Snajpera View Post
You are losing a lot of performance by disabling SMT! NEVER DO THAT ON AMD's CPUS!
Thanks for the advise. As far as I remember, I experienced exactly the opposite when using several encodes in parallel with pool feature and -- frame threads=1.
But I am going to stop my encodes to see if you are right indeed.


…..I couldn't prove that smt has more than no effect on fps on my System
maybe SMT is only useful in enoding with specific settings

Last edited by mparade; 24th June 2019 at 19:25.
mparade is offline   Reply With Quote
Old 24th June 2019, 20:20   #39  |  Link
Atak_Snajpera
RipBot264 author
 
Atak_Snajpera's Avatar
 
Join Date: May 2006
Location: Poland
Posts: 7,806
Quote:
Originally Posted by mparade View Post
Thanks for the advise. As far as I remember, I experienced exactly the opposite when using several encodes in parallel with pool feature and -- frame threads=1.
But I am going to stop my encodes to see if you are right indeed.


…..I couldn't prove that smt has more than no effect on fps on my System
maybe SMT is only useful in enoding with specific settings
Just run this benchmark without SMT and then with SMT
http://forum.pclab.pl/topic/1184884-x265-FHD-Benchmark/

If you disable SMT then you lose ~15%

Last edited by Atak_Snajpera; 24th June 2019 at 20:31.
Atak_Snajpera is offline   Reply With Quote
Old 24th June 2019, 22:07   #40  |  Link
mparade
Registered User
 
Join Date: Nov 2013
Posts: 577
I see but this 15% is not realized in fps during encode on my 1950X. No improvement is actually realized, just try it. If it is not the case by you, then your settings must be different than those of mine.

Last edited by mparade; 25th June 2019 at 07:26.
mparade is offline   Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 00:49.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.