QuEnc .71 VS HC .20

[Fishman0919]

I don't know how Procoder 2 and CCE can have almost the same speed.
There is some problem with your setup.
There is no encoder which is faster than CCE. With proper disk array and quad core PC (3.0Ghz) is about 10 times faster than RT.
Procoder 2 with "split encoding for multiple CPU" option used (can degrade quality) is at most 4 RT.

Not true, CCE still may be a bit faster even with encoding for multiple CPU on my dual core but there IS NO DEGRADE QUALITY

[Pulp Catalyst]

it's a shame that PSNR comparison is the only choice then,

perhaps it's time for a new method to be thought of that can look deeper into the image,

my thought would be comparing the original source using a mathematical formula with other versions done by other encoders & trying to reach 100% figure taking in as much information (as much as what can be conceived using mathematics) as possible,

i would of thought it would take time, and the mathematics needed would be extensive, also the analysis time would be extensive as well depending on how many tests would be done, and how many passes would be involved,


because clearly PSNR tests was not designed for this sort of thing,

p.s. isn't PSNR just for measuring noise or something like that, weird why many people do PSNR tests, i didn't think these tests meant anything unless one was looking for noise differences,

[techreactor]

it's a shame that PSNR comparison is the only choice then,



No perceptual tests are better as posted in my earlier post, and they take into consideration the human eye factor, but its more tedious and needs more people for the tests, therefore is not popular and put into practice.

[Fishman0919]

I did this test a long time ago, but still holds weight.


http://forum.doom9.org/showthread.php?p=624299#post624299

[techreactor]

I did this test a long time ago, but still holds weight.


Yeah! but the versions have changed since then (for the encoders) and so have their outputs/quality.

Perhaps it needs to get redone since its a perceptual test and depends on the visual judgement.

[kolak]

Not true, CCE still may be a bit faster even with encoding for multiple CPU on my dual core but there IS NO DEGRADE QUALITY


Sorry, but I'm lost.
I've said that CCE is the fastest available encoder and even if you use "Split encoding for multiple CPU (tested on 4 cores)" ProCoder 2 is still twice slower.

This option in ProCoder 2 may degrade quality, because every part is encoded with the same avg. bitrate and than they are joined together. You may be unlucky and one of the parts (which normaly would be encode with max. bitrate) will be encoded only with avg bitrate. It is more possible with short files.

[jdobbs]

What? :confused: That's not how DVD-RB works -- or am I missing something? Each segment is not encoded with the same average bitrate. The average bitrate is adjusted individually to keep the same proportional allocation as the original...

[Edit] When DVD-RB runs "Multiple Encoder Processes" there is a distinct instance of the software run for each encode... so the bitrates/settings can differ between them. The output for each segment would be bit-for-bit identical to what you would get if you ran a single instance -- so, as Fishman0919 mentioned, there would be no degraded quality. It would be exactly the same quality.

[Fishman0919]

Sorry, but I'm lost.
I've said that CCE is the fastest available encoder and even if you use "Split encoding for multiple CPU (tested on 4 cores)" ProCoder 2 is still twice slower.

This option in ProCoder 2 may degrade quality, because every part is encoded with the same avg. bitrate and than they are joined together. You may be unlucky and one of the parts (which normaly would be encode with max. bitrate) will be encoded only with avg bitrate. It is more possible with short files.


No, it doesn't degrade quality... dvd-rb tells 1 instance (we will say pro2 #1) to encode segment 1... and tells the 2nd instance (pro2 #2) to encode segment 2. It doesn't matter which one finishes first... it's told to start the 3rd segment... and so on. Instand of encoding one at a time. I'm not saying CCE isn't faster then 2 or 4 instances of Procoder 2... just that more then 1 instances of Procoder 2 at a time doesn't degrade quality

With my quad-core at work I get 4 instances of HC running, almost 25% faster then CCE Basic or CCE SP/SP2. Haven't had time to get any hard numbers.

EDIT: Damm jdobbs... I'm to slow again

PS. Some test with Procoder 3 run faster that CCE. And Quenc run about the same speeds as CCE on all my AMD tests... +/- a sec or 2

[kolak]

Sorry, but I was talking about ProCoder 2 as a standalone aplication (and its "Split encoding for multiple CPU@ option), not about dvd-rb.

You can run 2 instances of CCE, but you have to be logged in as a different user.

ProCoder 3 will use Rhozet engine which is optimised for multicore PCs, so it can beat CCE.
Is it already availabe?

[Sharc]

ProCoder 3 will use Rhozet engine which is optimised for multicore PCs, so it can beat CCE.


Beat CCE? What about one-pass encoding?

[kolak]

I don't know exactly, but I can imagine it can be as fast as CCE even for 1 pass encoding.
Rhozet plugin for Edius is really fast and uses 99% of all cores.

We will see...

[Fishman0919]

Form what I understand, Procoder 3 uses it's own Parallel-based encoding (from what the rep's at my company said)... similar to Rhozet... but based on Grid Encoding Technology from Rhozet Corporation.


Edit: LINK 1 (http://www.canopus.com/products/EDIUSPro/speedencoderforHDV_closeup.php) LINK 2 (http://www.canopus.com/products/ProCoderSW/newfeatures.php)

[Surf]

Quoting Hank's log with E6600 cpu:

"HC020
- 2 pass fast 30 min 4.32 GB
- 2 pass normal 31 min 4.32 GB
- 2 pass best 45 min 4.32 GB
- 1 pass fast/OPV 20 min 4.23 GB
- 1 pass normal/OPV 23 min 4.21 GB
- 1 pass best/OPV 32 min 4.20 GB"

45min max! "Jaw dropped"

CPU overclocked? OOOps, @3.16ghz... 04/22 feels like eons...

[kolak]

I've done some test and I didn't find HC encoder as good as CCE at all.
The source was some progressive animatiom from DigiBeta (uncompressed YUYV).

HC had many problem with flat areas (sky, etc.), there was lots of macroblocking, also problems on fades. I've used Best profile form HC gui with avg. 5Mbit and 8.5mbit max.

CCE looked sharper and had much less problems on flat areas (after some tweaking with CCE setings).

This was only on one perticular test file and with HC gui. I don't know if there is some cmd version with more settings to adjust.

Fishman0919 any coments ?

[gizzin]

Yes HC does have the problems with fade in and out scenes but I've only noticed that at lower bitrates 2500kb and under. The source average was 5mbit? At a bitrate of that high I doubt that there should be any macroblocking.

[kolak]

There were macroblockings on flat areas and
I didn't like it.

[jdobbs]

You might try changing the bias setting.

[Wodan]

I did my own little test with diff encoders,
it 2014 frames from Casino_Royale airport scene.
The Org m2v demuxed with DGIndex 1.49 Beta 16 is
5.858 Mbps size 60,248 KB. I reencoded for this test at 2500k and a max of 7500k.



GOP=12 (Auto for HC)
idct=9
Matrix=Mpeg2 standard (except AQE in 3 encodings)
Windows XP
AMD 64 3400+, 1g mem

I tested both CCE SP 2.70.02.12 and CCE SP2 1.00.00.13 but just put the results from CCE SP because all results were the same... Time, Size and Qual


Time Size PSNR Score
--------------------------------------------------------------------------------------------------------------
CCE Basic 2.70 | :59 sec| 25,588 KB | 45.2893
--------------------------------------------------------------------------------------------------------------
CCE SP 2.70 | | |
2-pass AQM | 1:00 | 25,611 KB | 45.3050
--------------------------------------------------------------------------------------------------------------
2-pass no AQM | :59 | 25,609 KB | 45.3133
--------------------------------------------------------------------------------------------------------------
3-pass AQM | 1:29 | 25,668 KB | 45.3217
--------------------------------------------------------------------------------------------------------------
3-pass no AQM | 1:29 | 25,687 KB | 45.3597
--------------------------------------------------------------------------------------------------------------
4-pass AQM | 1:59 | 25,676 KB | 45.3287
--------------------------------------------------------------------------------------------------------------
4-pass no AQM | 1:59 | 25,659 KB | 45.3535
--------------------------------------------------------------------------------------------------------------
10-pass no AQM | 4:59 | 25,669 KB | 45.3553
--------------------------------------------------------------------------------------------------------------
100pass no AQM | 50:46 | 25,669 KB | 45.3553
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
OPV Q-40 | :30 | 25,357 KB | 45.4022
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
HC .20 |
Best 2-pass | 1:47 | 25,627 KB | 45.7679
--------------------------------------------------------------------------------------------------------------
Normal 2-pass | 1:15 | 25,639 KB | 45.5196
--------------------------------------------------------------------------------------------------------------
Fast 2-pass | 1:11 | 25,632 KB | 45.3756
--------------------------------------------------------------------------------------------------------------
QuEnc .72 | | |
Nothing 2-pass | 1:00 | 25,028 KB | 45.3575
--------------------------------------------------------------------------------------------------------------
HQ | 1:29 | 24,190 KB | 45.3911
--------------------------------------------------------------------------------------------------------------
HQ Trellis | 2:32 | 24,271 KB | 45.1942
--------------------------------------------------------------------------------------------------------------
HQ Trellis | | |
Extreme | 4:58 | 24,154 KB | 45.2267
--------------------------------------------------------------------------------------------------------------
HQ Extreme | 3:47 | 24,800 KB | 45.4234
--------------------------------------------------------------------------------------------------------------
AQE | | |
No QMatOp | 1:26 | 25,692 KB | 45.6325
--------------------------------------------------------------------------------------------------------------
Full QMatOp | 3:14 | 25,629 KB | 44.0711
--------------------------------------------------------------------------------------------------------------
Mod QMatOp | | |
max_intra=95 | | |
max_inter=85 | 3:13 | 25,630 KB | 44.2795
--------------------------------------------------------------------------------------------------------------
Mod QMatOp | | |
max_intra=85 | | |
max_inter=45 | 3:13 | 25,628 KB | 44.5490
--------------------------------------------------------------------------------------------------------------
Procoder 2 | | |
Master Qual | 1:57 | 25,636 KB | 45.6491
--------------------------------------------------------------------------------------------------------------
Highest Qual | 1:50 | 25,635 KB | 45.6382
--------------------------------------------------------------------------------------------------------------
High Qual | 1:31 | 25,643 KB | 45.1911
--------------------------------------------------------------------------------------------------------------
Speed Qual | 1:29 | 25,641 KB | 44.8248
--------------------------------------------------------------------------------------------------------------
TMPGEnc 4.0 | | |
Highest with | | |
error correction | 3:26 | 25,573 KB | 44.6268
--------------------------------------------------------------------------------------------------------------
Standard | 3:19 | 25,577 KB | 44.4551
--------------------------------------------------------------------------------------------------------------
MainConcept | | |
Default | 1:36 | 25,649 KB | 43.4772
--------------------------------------------------------------------------------------------------------------

100 pass test still in progess....DONE


Edit: Added OS, CPU, notes and fixed spelling


like dragongodz said (post below)... do not take metric scores as absolutes... just thought I would add it seeing that I had all this .m2v files from the test.

sorry to bump this, but thats pretty impressive by hc encoder...looks like it beats cce. Will give hc encode a shot and see how it does:thanks: