OBS branch with AMD VCE support.

Anijatsu

New Member
I was unable to find any posts concerning if it's possible to mitigate blockiness of bright colours. I guess this has something to do with colour space conversion. Is it something that's caused by hardware? I tried both OVE and AMF, it's the same thing.
2015-05-04_16-37-37.png
2015-05-04_16-37-11.png
 

khades

Member
Looking to buy radeon card for streaming, i cant really find any information about amd vce quality for streaming and most of tests i saw used extreme bitrate settings (like 35k)

i have several questions

1) How good amd vce 2.0 compared to haswell's quicksync?
2) How good amd vce 2.0 compared to x264 veryfast preset?
3) Is there any benefit of taking r9 285 over r9 290 for encoding (i know 285 is inferior by performance, but its vce block is faster)?
4) What's optimal bitrate for streaming with vce for 720p30, 720p60, 1080p30, 1080p60?
 
Last edited:

dping

Active Member
The standard mode works fine but I want to be able to name files and have the container as mp4 beforehand so I can keep track of recordings more easily and to also save me from having to remux (which for me takes a while as a lot of my clips are 1hr+) the clip afterwards to use in Sony Vegas. It needs to be using VCE AMF because the default x264 encoding lags me out tonnes whereas VCE AMF allows me to record 1080p 60fps no problem
@jackun, From what I am hearing, remuxing to AVI (default FLV) does not work with AMF. Any thoughts? btw, @ZombieWizzard What version of OBS MP (VCE fork) are you using?

Looking to buy radeon card for streaming, i cant really find any information about amd vce quality for streaming and most of tests i saw used extreme bitrate settings (like 35k)

i have several questions

1) How good amd vce 2.0 compared to haswell's quicksync?
2) How good amd vce 2.0 compared to x264 veryfast preset?
3) Is there any benefit of taking r9 285 over r9 290 for encoding (i know 285 is inferior by performance, but its vce block is faster)?
4) What's optimal bitrate for streaming with vce for 720p30, 720p60, 1080p30, 1080p60?
I cannot answer you VCE vs QSV in the contact of a version by version standpoint but in general, VCE 1.0/2.0 are about the same quality/speed with the exception of b frames, which have been added to VCE 2.0.

As for the r9 285, Yes, it has an encoder that can technically do up to 4k streaming, but we've only had one guy with one and was unable to test anything higher than 1080@60 since I believe that there is spotty support for this in general. That being said, if the encoder can do 4k encode, that means that it should be able to do 1080@60 using quality preset as long as there isn't code added to block this (which I believe there is); if it will work, it will be able to do it faster than an r9 290 (VCE 2.0) card in general.


VCE in general is about equivalent between superfast and very fast preset so about a 500 bitrate will be needed for equivalent quality.
 
Last edited:

khades

Member
VCE in general is about equivalent between superfast and very fast preset so about a 500 bitrate will be needed for equivalent quality.

Ok, so will i be able to get a decent quality of picture with 3500kbit at 1080p30 or 720p60 at vce 2.0 videocards?

I currently have haswell chip and use quicksync and it is blocky at 720p30 with 2400 bitrate, and it seem to lack a lot of details, that might be preserved by b-frames. I am really not impressed by quicksync at all and i really want to know if vce is better
 

dping

Active Member
Ok, so will i be able to get a decent quality of picture with 3500kbit at 1080p30 or 720p60 at vce 2.0 videocards?

I currently have haswell chip and use quicksync and it is blocky at 720p30 with 2400 bitrate, and it seem to lack a lot of details, that might be preserved by b-frames. I am really not impressed by quicksync at all and i really want to know if vce is better
720@30 needs 2500bitrate for best with x264 so with any hardware decoder, you'll need more, maybe 3000.
 

dping

Active Member
Guys if i am going to upload to youtube at 1080p60 what would the recommended gop and idr period-?
GOP 30 IDR 120
set the quality preset to balanced or speed preferably if balanced stutters at all. the speed preset wont lose any noticable quality as long as there is enough bitrate to compensate.

I would also set your bitrate a little higher VCE's max for 1080@60 is 50 000
 

Krelian

Member
GOP 30 IDR 120
set the quality preset to balanced or speed preferably if balanced stutters at all. the speed preset wont lose any noticable quality as long as there is enough bitrate to compensate.

I would also set your bitrate a little higher VCE's max for 1080@60 is 50 000

Thanks, but i am setting the bitrate to vbr 25, 30 top, i know the max for vce is 50 but for some reason youtube encoder degrades it a lot past 35-40. My personal tests have led me to believe just as 10 is the max for 1080p30, 35-40 is the max for 1080p60.
 
Here is an old comparison one made for me between Haswell and VCE 1.0:
http://ohkd.screenshotcomparison.com/comparison/78902

Worth noting is that the bitrates weren't identical, i think both were around 20-30mbps, i can't remember which had more bitrate, might have been VCE.

However the difference even if you calculate the Bitrate was huge between them (as you can see).
Haswell is much worse than VCE.

Worth noting is that i also compared it to x264 Superfast, and it looks quite a bit better if i can remember correctly.
So VCE is not between Superfast - Veryfast but rather Superfast - Ultrafast.

(Worth noting is that if you tweak settings results may vary, this was a random quick test i asked for).

EDIT: looked up the old posts , apparently they were same Bitrate in the 26mbit version, it was just that 50mbps didn't work with the AMD. (If i remember correctly the 50mbps QS looked almost the same as 26mbps VCE).
http://forum.doom9.org/showpost.php?p=1683567&postcount=25
 
Last edited:

dping

Active Member
Here is an old comparison one made for me between Haswell and VCE 1.0:
http://ohkd.screenshotcomparison.com/comparison/78902

Worth noting is that the bitrates weren't identical, i think both were around 20-30mbps, i can't remember which had more bitrate, might have been VCE.

However the difference even if you calculate the Bitrate was huge between them (as you can see).
Haswell is much worse than VCE.

Worth noting is that i also compared it to x264 Superfast, and it looks quite a bit better if i can remember correctly.
So VCE is not between Superfast - Veryfast but rather Superfast - Ultrafast.

(Worth noting is that if you tweak settings results may vary, this was a random quick test i asked for).

EDIT: looked up the old posts , apparently they were same Bitrate in the 26mbit version, it was just that 50mbps didn't work with the AMD. (If i remember correctly the 50mbps QS looked almost the same as 26mbps VCE).
http://forum.doom9.org/showpost.php?p=1683567&postcount=25
Yeah, I've been saying between super and veryfast for a while now. do you think you could do your test with the current AMF? one for each speed/balanced/quality. Maybe just do a test for 720@30 15000. or maybe I'll do one when I get home, but I wont be able to test QSV...
 
Sadly i wasn't the one doing those tests as i don't have AMD VCE, i am as interested in the results as you are:(
Especially interested how R9 285 is, cause if they improved the quality a bit more, it would clearly Superfast which would be splendid. (Though perhaps they already are there with Quality preset, i do not know).
 

Kobata

Member
Sadly i wasn't the one doing those tests as i don't have AMD VCE, i am as interested in the results as you are:(
Especially interested how R9 285 is, cause if they improved the quality a bit more, it would clearly Superfast which would be splendid. (Though perhaps they already are there with Quality preset, i do not know).
Here's a quick test: Google drive (download the videos, don't use the embedded viewer)

I wasn't entirely thinking, so I forgot to change the resolution down when recording the source, but it's letterboxed so cropping down to 1080p for VCE1/2 tests shouldn't be a very major issue.

The screenshots are from frame 350, and the VCE ones look fairly bad (possibly because of differences in keyframe placement as I didn't give any settings for that) -- in motion it seems fairly good, clearly better than ultrafast, possibly a little worse than superfast.
 

dping

Active Member
Hi I have been testing VCE since yesterday. I don't know why but I have been dropping a lot of frames for no reason. My GPU load wasn't even near 100% and my internet is more than fine.

Been trying to stream at 720p 60fps

https://gist.github.com/7c930cb10506226687e3
What game? have you played with toggling dx10 interop?

I dont know if that was due to VCE or not but:

Code:
Number of times waited to send: 6094, Waited for a total of 22730599 bytes

above usually means that the connection was stalling out, but again, it could be the encoder but I only see 6 frames (0.02%) dropped due to encoder lag...

EDIT: your ingest also took 1300ms to connect. can you try the same settings via x264?
 

Pharaphobia

New Member
Hi, thx for fast replay..

When I stream with x264 there are 0 dropped frames, but it could be due to bad server (Prague is known to be total crap)..

The game i stream is World of Warcraft (64bit client)

P.S: Is my settings actually okay? the advanced settings has way to much unknown information for me :D
 

dping

Active Member
Hi, thx for fast replay..

When I stream with x264 there are 0 dropped frames, but it could be due to bad server (Prague is known to be total crap)..

The game i stream is World of Warcraft (64bit client)

P.S: Is my settings actually okay? the advanced settings has way to much unknown information for me :D
lately, the LOG is difficult to read as for some settings. like whats your GOP set to? IDR? are you usings host or dx9 or dx10?
 
Top