TMPGenc 2.52 plus: The end of CQ_VBR!

[SatStorm]

Here are my conclusions of the new TMPGenc 2.52 plus, after some tests I made this weekend:

With the new 2.52 plus version, using 2 Pass VBR (new ) with "time estimate search (fast)", you get much better results with mpeg - 2 files, than using CQ_VBR with "highest quality (very slow), AT THE SAME TIME
So, in a way we have a speed increase, because of the quality increase!
With my athlon xp 1700+, setting 2 Pass VBR to "time estimate search" or CQ_VBR to "highest quality" the speed is about 0.5X for encoding any source to 352 X 288 mpeg 2 and about 0.33 for encoding any source to 352 X 576 mpeg 2.

So, it is really, really, really useless to work with CQ_VBR anymore!

For those who like to test, here some settings:

For 352 X 288 use the following settings:
2Pass VBR: min 220, average 1150 max 2300. Time Estimate Search (fast)
CQ_VBR: min 220, max 1340, quality: 83%. Highest Quality (very slow)

For 352 X 576 use the following settings:
2Pass VBR: min 220, average 1900, max 2600. Time Estimate Search (fast)
CQ_VBR: min 220, max 3000, quality: 85%

compare file size/quality/time
2 Pass VBR is a clear winner. The same time for encoding with smaller files and much much much better quality!

For the very first time I saw NO blocks at all! It is like CCE 3Pass VBR...


For the record:
Setting 2 Pass VBR to "high quality (slow)", you get even better results, with extreme moving scenes. For the 95% of all cases, you don't need to use "high quality". Time Estimate is enough!
I saw no difference between "high quality" and "highest quality". Don't use highest quality, it's just a waste of time....

goodbye CQ_VBR! It was nice using you the last year!

[sue denim]

I couldn't agree more - the 'new' 2Pass VBR is superb!!

The only thing I've got against TMPGenc now is that it's bit-distribution is still a way off being perfect ..... file sizes are huge compared to CCE.

I can get around 70mins of sharp, blockless MPEG2 video/audio on an 80min CDR with CCE ........ compared with around 55mins with TMPGenc!!

Sue xxx

[plg-tae]

faggets...cq_vbr is for lazy fucks with slow computers...explain it in any way u like,,,multipass has always been the way to go...you are saying this now cause you finally got a decent cpu,,not cause tmpeg changed,,,cq_vbr is merely 1 pass vbr..so of course 2 pass or more is better..especially with cce...u can reply to this all you want...you are a dumb ass and i wont reply to this any further....jackasses

[sue denim]

hmmmmm ....... this guy's very eloquent, isn't he SatStorm?!!!?

......... as subtle as an ice-dancing elephant ....... (and then some....) .....

It's true, I have just bought myself a new Athlon box but I still stand by my comment - the 'new' 2Pass in the PLUS versions beats the 'old' by a massive margin.

Can't wait for the server version ...... It'll probably cost an arm and a leg though!

Laters,

Sue xxx

[SatStorm]

@plg-tae: You are a CCE owner? You are rich! I am not. That is the difference. So, I am very happy, for my 40$ Tmpgenc....
A 3960$ difference, easily explain the difference between CCE and TMPGenc I think!

About 2Pass VBR vs CQ_VBR:
CQ_VBR with Highest quality create worst files at the same time as 2Pass VBR with Time Estimate Search, at any CPU.
If a 4 min clip needs 20 min to encode with CQ_VBR at highest quality mode with a duron 900 for example, it takes 20 min to encode with 2Pass VBR at Time Estimate Searh mode. The file produced by 2Pass VBR is smaller, better looking and , block free with normal bitrates.

Yesterday I made more tests!
The new 2Pass VBR is amazing: 352 X 576 with min 220,average 1900, max 2520 is totally block free! The quality is identical Cinemacraft with the same testings. Speed is about 20% slower.
Not bad for 40$....

[Martí]

I can't wait to try it! Will post my opinion too!!

Martí

[alfa1234]

faggets...cq_vbr is for lazy fucks with slow computers...explain it in any way u like,,,multipass has always been the way to go...you are saying this now cause you finally got a decent cpu,,not cause tmpeg changed,,,cq_vbr is merely 1 pass vbr..so of course 2 pass or more is better..especially with cce...u can reply to this all you want...you are a dumb ass and i wont reply to this any further....jackasses

Is your English style a fad or merely a show of uneducated language ignorance? Perhaps some respect to your friends and co-writers might improve your stupid image.
Live and learn!

[g0dZpr0fi7]

plg-tae
I agree with what you said but not the way you said it *******!

[TheSeven7hSon]

"i wont reply to this any further...."

Good.

[Chips144]

Let me address first things first. I realize that it would be impossible for moderators to read all incoming messages before being posted. However, if the only thing you can write is full of (******, *******, etc.) with absolutely no constructive criticism or added information, then perhaps this forum would be better off without you (plg-tae). Perhaps our moderators would consider membership deletion for this type of contributor. At least that way he/she would have to obtain a new e-mail address before he/she could once again interupt the flow of useful information.
Now for the real issue. I have a P4 1.7GHz, running WindowsXP and TMPGEnc 2.52 Plus. I ran an experiment to see the difference between CQ_VBR and 2 pass VBR. I am encoding to SVCD. For this test I choose a 1 minute 37 second clip from the DVD "Wolf". I choose this particular clip because it contains dark scenes when Jack nicholson hits the wolf while driving at night and then moves to a bathroom steam-on-the-mirror scene to a scene where Jack is walking past a complex construction site. He is walking behind a chain-linked fence and the camera is panning at the same time. This just about encompasses all the worst case senarios you have to deal with in encoding. I ripped with Smartripper and used DVD2AVI with "Forced film" set (the DVD is encoded progressive) to frameserve to TMPGEnc. Here are the important settings and results:

TEST 1
CQ_VBR
max = 2520
min = 320
Q = 70
Motion Search Precision = High Quality (slow)
Time to encoide = 4:55
Encoded file size = 31.38Mb

TEST 2
2 pass VBR
max = 2520
ave = 1900
min = 320
Motion Search Precision = Motion Estimate Search (fast)
Time to encode = 6:15
Encoded file size = 25.46Mb

TEST 3
2 pass VBR
max = 2520
ave = 2100
min = 320
Motion Search Precision = Motion Estimate Search (fast)
Time to encode = 6:17
Encoded file size = 27.73Mb

TEST 4
2 pass VBR
max = 2520
ave = 2100
min = 320
Motion Search Precision = Normal
Time to encode = 6:42
Encoded file size = 27.72Mb

TEST 5
2 pass VBR
max = 2520
ave = 2100
min = 320
Motion Search Precision = High Quality (slow)
Time to encode = 9:23
Encoded file size = 27.72Mb

After buring each test onto separate CD-R's, I played them through an Apex 5131 onto a Toshiba 65-inch HDTV. I purposly mixed them up before playing in order to perform a "blind" test. The results were amazing. THEY ALL LOOKED THE SAME. I could not tell which had been encoded using 2 pass. Each individual has their own priorities. Mine are QUALITY, TIME and lastly FILE SIZE (within reason). Since I could not see any difference in quality, I would choose CQ_VBR because the time to encode is shorter. Obviously if you choose Motion Search Precision = Highest Quality (very slow), the numbers change. However, every post I have read on this setting said not to use it because it simply wastes time and no significant improvement can be seen. Has anyone else seen this type of performance using CQ_VBR? In truth, for what I see, for the quality to improve beyond this you have to have the actual DVD. Which I do have and play the original through a Toshiba SD6200 Progressive Scan DVD player as my benchmark. These results seem to contradict those of most other posts on this subject. Does anybody have any thoughts?

[pepehdez]

I have a question, where can I get TMPGenc 2.52 plus.

[SatStorm]

That was a very good test you made Chips144
But your source was DVD/progressive and the output was progressive I believe...
There is a huge speed difference between progressive/interlace output. TMPGenc, generally encodes faster when the output is Interlace...
Also, CQ_VBR interlace output produce blocks. Those blocks don't appear on progressive output.
Anyway, my source is DVB transmissions, all interlaced, some times full PAL (720 X 576) but most of the times 544 X 576. I also do many tests with VHS source...
My tests are Basicly Music Video Clips. I have some really difficult ones to test. For example, the 1988's video "The right stuff" from Bryan Ferry. Full motion all the time, many close ups etc...
The greatest terror of CQ_VBR is the camera's close ups! CQ_VBR never creates blockfree mpeg 2 of a moving head. Those who encodes Heavy Metal or Early 90s Techno Rave videos can understand what I mean. The moving hair is the horror... Colorfull Backrounds also! The only solution with CQ_VBR is more bitrate. I don't like this solution.
On typical movies, I can encode up to 105 min to one CD using CQ_VBR with impressive results! But it is not the same with music videos...
Also, tmpgenc 2.52 is generally optimised for Athlon xp and Pentium 3 !!!! There is no improvement with p4 as people report...
The encoding time with my athlon xp 1700, for a clip let say 3 min, to 352 X 288 mpeg 2 interlace, is about 6.20min with CQ_VBR and 6.30 with 2Pass VBR Motion Search Estimate mode.
The encoding of the same clip to 352 X 576 mpeg 2 interlace is about 12 min both CQ_VBR (highest) and 2 Pass VBR (MSE)
The difference is huge. 2 Pass has no blocks! Amazing! And in the same encoding time...
Maybe my mobo is good, what to say...
But I don't use CQ_VBR anymore, it gives me nothing!

[Chips144]

Thanks SatStorm for your reply. You are correct, I am encoding to progressive output. Your explanation of interlaced vs. progressive output in TMPGEnc I'm sure is the reason I get the results I reported. I also have not seen any difference in encoding speed when going from a P3 to a P4 except that which would be expected by the increase in processor speed, i.e. 700MHz to 1.7GHz and going from TMPGEnc 2.02 to 2.52 Plus.

[Rich927]

Intresting....We have a Forum here which we can post information to and learn from. However, some people feel they have to trash and degrade others with the use of profanity. I for one, would like to see those who do this barred from posting. Some of us may not know as much as others and are here to learn. If I could afford it, I would be using Pro level hardware and software. Alas, that is not the case so, I limp along with my PII 450, Sony TRV8, VideoStudio, Pinnacle Studio and TMPgenc. I'm having fun and learning. The day I have to resort to using profanity, I'll consider myself STUPID!..... The only dumb question is the one not asked. Profanity is the first sign of ignorance.

[R_A_P_D_]

Hi, I just bought Tmpgenc 2.52 plus but there is no "time estimate search (fast)" in 2 pass VBR...

Rather in the Video tab screen under Motion search precision
there are 6 choices:

lowest quality (Very fast)
low quality (fast)
Normal
High quality (slow)
Highest quality (very slow)
Motion estimate search (fast)

Do you mean "Motion estimate search (fast)"????

Thanks for help!

[SatStorm]

@R_A_P_D: Yes, I mean Motion estimate search (fast)...
Sorry for the misunderstanding!