umh vs. subme 6 vs. subme 7

[roozhou]

And even if the speed loss for p4x4 is smaller than the speed loss for HEX -> UMH, this isn't really an argument. UMH isn't a default either ;)

(BTW: I think your test with the 360p version of Big Buck Bunny may be biased, because p4x4 is more helpful for "lower" resolution video, if helpful at all)

A rank of performance loss: p4x4 < hex->umh < m6->m7

So generally which one brings more quality, umh or subme 7?

[Audionut]

So generally which one brings more quality, umh or subme 7?

For the purpose of x264 defaults, you also need to consider the speed loss of each option.

edit, and now you've just completely changed the subject.
At the end of the day, I think it's fairly safe to consider that the x264 devs want to use the best* options for defaults.
I think it's also fairly safe to consider that the x264 devs have considerably more experience and users to base their decisions on.
I also believe that most of your posts bash the decisions of the x264 devs. And so the point of you trolling still is valid.

*sorry

[roozhou]

At the end of the day, I think it's fairly safe to consider that the x264 devs want to use the best* options for defaults.
I think it's also fairly safe to consider that the x264 devs have considerably more experience and users to base their decisions on.

I suggest x264 devs add umh to default, or remove subme 7 from default.

A hex+m7 vs umh+m6 test will show which one has higher quality/performance ratio.

[Audionut]

A hex+m7 vs umh+m6 test will show which one has higher quality/performance ratio.

I've got the weekend off. Expect some results.

edit: probably not in your favour, but meh.

[roozhou]

With B-frames to be on by default it makes sense to use subme 7 so that RD is performed on all frames. Plus Dark Shikari has said that he would prefer to raise subme before raising me.

Don't forget psy-rd. Subme 7 may bring worse quality if psy-rd is too high.

[LoRd_MuldeR]

Don't forget psy-rd. Subme 7 may bring worse quality if psy-rd is too high.

But you really can't blame Subme 7 for that. If the chosen Psy-RD settings are bad for a source at Subme 7, they were bad at Subme 6 as well!

Subme 7 just makes the problem more visible, as it runs RD (and hence Psy-RD) on more frame types.

On could even argue that Subme 7 makes it easier for average Joe to identify improper Psy-RD settings, compared to Subme 6 ;)

[roozhou]

But you really can't blame Subme 7 for that. If the chosen Psy-RD settings are bad for a source at Subme 7, they were bad at Subme 6 as well!

Subme 7 just makes the problem more visible, as it runs RD (and hence Psy-RD) on more frame types.

On could even argue that Subme 7 makes it easier for average Joe to identify improper Psy-RD settings, compared to Subme 6 ;)

Once again you missed my point. Does umh has similar problem? If not why exclude it from the default? It can never make things worse.

No one will determine whether psy-rd is good by comparing m6 and m7. It is practically quite difficult to find the exact value for psy-rd with which subme 6 and subme 7 will have identical visual quality. So using subme 6 as default would minimize the potential quality loss of improper psy-rd setting.

[LoRd_MuldeR]

That's a weird way of thinking. If the value of parameter X is bad for your source, then the only proper way to fix this is correcting the value of parameter X (or turn off X completely). What you suggest is: Hey, lets keep parameter X at its obviously improper value, but lower parameter Y instead, because with Y lowered X will have less chances to show its negative effect -- which it still will have, just in fewer cases. So instead of fixing the problem, you suggest to hide the symptom. That's a kind of hackish workaround. And it will never look as good as Subme 7 with a suitable Psy-RD strength...

[roozhou]

Hey, lets keep parameter X at its obviously improper value
Obviously? There isn't an easy way to tell whether the value of psy-rd is proper.

If you have one, please let me know.

[LoRd_MuldeR]

Obviously? There isn't an easy way to tell whether the value of psy-rd is proper.

If you have one, please let me know.

Subme 7 looking worse than Subme 6 is an indication of a bad Psy-RD value, I'd say ;)

With a proper Psy-RD value (or with Psy-RD disabled) Subme 7 should never look worse than Subme 6.

And in the unlikely case Subme 6 really looks better than Subme 7 (without Psy-RD), you can get even better quality with Subme 5 :D

Whatever Subme 7 does bad (if anything), Subme 6 does bad as well. Just on fewer frames...

[Audionut]

There isn't an easy way to tell whether the value of psy-rd is proper.

If you have one, please let me know.

You have one just to the left of your nose, and another one just to the right of your nose.

[roozhou]

Subme 7 looking worse than Subme 6 is an indication of a bad Psy-RD value, I'd say ;)

With a proper Psy-RD value (or with Psy-RD disabled) Subme 7 should never look worse than Subme 6.

And in the unlikely case Subme 6 really looks better than Subme 7 (without Psy-RD), you can get even better quality with Subme 5 :D

Whatever Subme 7 does bad (if anything), Subme 6 does bad as well. Just on fewer frames...

Why are x264 devs planning to change defaults? Because it brings convenience.

Those who use defaults will never have time and patient to do such comparison. If you insist on keeping subme 7 as default, you should also lower the default psy-rd value so that subme 7 being worse than subme 6 can never happen.

[LoRd_MuldeR]

If the default Psy-RD strength really behaves "bad" on most clips (and that is independent from Subme 6 vs. 7), then it should be lowered indeed. But only if!

It should not be lowered, just because it behaves bad on a few clips, as long as it helps on the majority of clips. Defaults always are a trade-off ;)

Also note that there will be a "--tune" option along with the new "--preset" option. And "--tune" adjusts Psy-RD based on the type of content (Film vs. Anime).

[roozhou]

It should not be lowered, just because it behaves bad on a few clips, as long as it helps on the majority of clips. Defaults always are a trade-off ;)


Here is my point. UMH never behaves worse than HEX. It helps on all clips. Why exclude it from defaults?

[Audionut]

Here is my point. UMH never behaves worse than HEX. It helps on all clips. Why exclude it from defaults?

UMH is probably a bit too costly speed-wise to set as the default.

As I have said before.

For the purpose of x264 defaults, you also need to consider the speed loss of each option.

Some could very well argue that UMH does indeed behave worse than HEX because it's slower.

[LoRd_MuldeR]

Here is my point. UMH never behaves worse than HEX. It helps on all clips. Why exclude it from defaults?

Because UMH is too slow for the default (aka "--preset medium"). It will be included in the slower presets. Sounds reasonable too me ;)

Also to make this clear again: Subme 7 never behaves worse than Subme 6. If it ever would (very unlikely), then Subme 5 would be even better. Yes, Psy-RD with default strength may behave bad on some sources, simply because the default strength can't be optimal for any source. And in that case the problem (which originates from improper Psy-RD strength), becomes more visible at Subme 7 than on Subme 6. But you rally can't conclude anything about Subme 6 vs. Subme 7 from that. Whatever Subme 7 does bad (if anything), Subme 6 does bad as well! Just on fewer frames...

[roozhou]

UMH too slow? Subme 7+HEX is slower than subme 6+UMH.

@ Mulder

Because UMH is too slow for the default (aka "--preset medium").

Subme 7 never behaves worse than Subme 6.
None of these statement are based on real tests. One test sample is more convincing than a thousand words.

Whatever Subme 7 does bad (if anything), Subme 6 does bad as well!
How can you determine whether one clip is "bad"? We are talking about "better" and "worse". When the bitrate is low enough the result is always "bad", but it doesn't stop me from getting a "better" result at such bitrate.

[LoRd_MuldeR]

UMH too slow? Subme 7+HEX is slower than subme 6+UMH.

Too slow for the improvement it delivers, of course! Simply being slower or faster doesn't mean anything.

Obviously the x264 devs are sure that "Subme 7 + HEX" gives more improvement over "Subme 6 + HEX" than "Subme 6 + UMH" would give.

Therefore Subme 7 is worth the slow down it causes, UMH isn't. Not at "medium" preset at least...

None of these statement are based on real tests. One test sample is more convincing than a thousand words.

It's a simple matter of logic:

If a < b and b < c then a < c :p

How can you determine whether one clip is "bad"? We are talking about "better" and "worse". When the bitrate is low enough the result is always "bad", but it doesn't stop me from getting a "better" result at such bitrate.

To explain it once again :rolleyes:

The difference between Subme 5 and Subme 6 is that Subme 6 runs RD on I- and P-Frames. The difference between Subme 6 and Subme 7 is that Subme 7 runs RD on all frame types (including B-Frames). So if Subme 6 looks better than Subme 5, then RD improves quality. And consequently Subme 7 looks even better, because it runs RD on more frame types than Subme 6. If Subme 7 ever looks worse than Subme 6, this would mean RD hurts quality! But then RD would hurt quality in Subme 6 the same way it hurts quality in Subme 7! This leads to the one and only conslusion: If Subme 7 ever looks worse than Subme 6, then Subme 6 also looks worse than Subme 5 too. If you ever think that Subme 7 gives worse result than Subme 6, don't switch to Subme 6, switch to Subme 5 !!!

[burfadel]

Or Subme 9 :D

[roozhou]

The difference between Subme 5 and Subme 6 is that Subme 6 runs RD on I- and P-Frames. The difference between Subme 6 and Subme 7 is that Subme 7 runs RD on all frame types (including B-Frames). So if Subme 6 looks better than Subme 5, then RD improves quality. And consequently Subme 7 looks even better, because it runs RD on more frame types than Subme 6. If Subme 7 ever looks worse than Subme 6, this would mean RD hurts quality! But then RD would hurt quality in Subme 6 the same way it hurts quality in Subme 7! This leads to the one and only conslusion: If Subme 7 ever looks worse than Subme 6, then Subme 6 also looks worse than Subme 5 too. If you ever think that Subme 7 gives worse result than Subme 6, don't switch to Subme 6, switch to Subme 5 !!!

You don't have to explain why subme 7 may look worse than subme 6. I know it.

How can you know whether subme 6 is worse than subme 7? You need two 2-pass encodings at the same bitrate. You cannot make the conclusion "the result is bad, I should either switch to subme 5 or lower psy-rd" by just encoding a few frames with default settings.

So the problem is you cannot ask those who use default settings to do such comparison. We should include settings with best quality/performance gain into default and exclude settings or setting combinations which may bring both worse quality and slower encoding speed.

[LoRd_MuldeR]

How can you know whether subme 6 is worse than subme 7?

I think nobody would really doubt that Subme 6 looks better than Subme 5. If thas wasn't the case, RD would be useless and should be removed :p

But if Subme 6 looks better than Subme 5 (that is: RD does improve quality), then Subme 7 looks better than Subme 6 by definition. Or in other words: Subme 6 looks worse than Subme 7.

Since this discussion is stuck and I can only repeat what has been said in this thread, I will stop at this point ;)

[roozhou]

Or in other words: Subme 6 looks worse than Subme 7.


Yes, you need to LOOK at both of them, that means you need to ENCODE both of them and you lose the convenience of using default.

[froggy1]

Yes, you need to LOOK at both of them, that means you need to ENCODE both of them and you lose the convenience of using default.

dude, stop trolling about subme 6 and 7. As DS said in the past, you're a troll and I'd like to warn the others to not fall in your trolltrap. I wouldn't be surprised that next time you'll come claiming that subme 4 looks better than subme 7 or 8..

[roozhou]

I wouldn't be surprised that next time you'll come claiming that subme 4 looks better than subme 7 or 8..
No, it won't happen.

stop trolling about subme 6 and 7
I am comparing subme 7 and umh, not subme 6 vs subme 7.

And if i am trolling.
UMH is too slow for the default
So is he.

[roozhou]

Would some mod be so nice to move all the Troll posts to a separate thread and then close that thread please?

In the hope that this thread can focus on the initial topic again...

I suggest including UMH into default. But you guys argued subme 6 vs subme 7.

[lansing]

I suggest including UMH into default. But you guys argued subme 6 vs subme 7.

oh god, you should really enhance your logical thinking, and common sense.

You are arguing that UMH+subme 6 > HEX+subme 7 in terms of speed/quality.

and the simple FACTs:
subme 7 is better than subme 6;
UMH is way slower than HEX

And then the logic goes:
a way faster HEX search with a better quality subme 7 > a lot slower UMH search with a lesser quality subme 6.
Simple and clear.

and here's the logic for why you're argument about subme 6 > subme 7 cannot be true:

If you deny the FACT that subme 7 is better than subme 6 in any means, then you're denying the fact that subme 9>8>7>6... hence, you're saying that the whole subme calculation was wrong.:eek: Use your common sense and think like a normal human being man.

[neuron2]

@lansing

You're struck for rule 4.

@roohzou

Please do not respond to the provocations. Thank you.

@all

We can disagree civilly about things. There's no need to get nasty.

[roozhou]

If a lot of b-frames are used, UMH + subme 6 runs faster than HEX + subme 7. We all know why.

I did plenty of tests. On regular film samples HEX + subme 7 is ~10% faster than UMH + subme 6, but on animes or video with a lot of fades UMH + subme 6 are faster.

[Sharktooth]

but anime and lot of fades are not "standard" content... they're just case scenarions. plus the tune option will take care of animated content... so it's a matter of personal preference: you want those options set that way coz they would be optimal to you and some anime encoders...
hence, i dont think it's a good idea.

[Audionut]

Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 1 --bitrate 2000 --stats "F:\subme6.stats" --level 4.1 --keyint 240 --min-keyint 24 --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --subme 2 --partitions none --me dia --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output NUL "F:\test.avs"
--[Information] [4/07/2009 8:09:37 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile Main, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:21.08 size: 56017
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.20 size: 13289
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.85 size: 3966
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 46.3% 0.0% 53.7%
---[NoImage] x264 [info]: mb P I16..4: 17.3% 0.0% 0.0% P16..4: 50.2% 0.0% 0.0% 0.0% 0.0% skip:32.6%
---[NoImage] x264 [info]: mb B I16..4: 0.9% 0.0% 0.0% B16..8: 21.5% 0.0% 0.0% direct:12.0% skip:65.6% L0:26.4% L1:48.6% BI:25.0%
---[NoImage] x264 [info]: final ratefactor: 21.40
---[NoImage] x264 [info]: direct mvs spatial:98.4% temporal:1.6%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:33.9% 44.3% 12.2% inter:18.1% 12.0% 0.8%
---[NoImage] x264 [info]: kb/s:1926.3
---[NoImage] encoded 10001 frames, 88.61 fps, 1926.42 kb/s
Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 2 --bitrate 2000 --stats "F:\subme6.stats" --me hex --subme 6 --level 4.1 --keyint 240 --min-keyint 24 --ref 3 --mixed-refs --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output "F:\subme6.mkv" "F:\test.avs"
--[Information] [4/07/2009 8:11:43 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile High, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:20.63 size: 59818
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.02 size: 13604
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.29 size: 4277
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 11.8% 69.9% 18.3%
---[NoImage] x264 [info]: mb P I16..4: 1.2% 3.6% 0.6% P16..4: 45.7% 11.2% 11.9% 0.0% 0.0% skip:25.7%
---[NoImage] x264 [info]: mb B I16..4: 1.5% 0.0% 0.0% B16..8: 24.0% 1.4% 1.2% direct:10.8% skip:61.2% L0:32.3% L1:45.5% BI:22.2%
---[NoImage] x264 [info]: 8x8 transform intra:59.7% inter:51.8%
---[NoImage] x264 [info]: direct mvs spatial:87.0% temporal:13.0%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:60.5% 68.8% 34.5% inter:16.6% 18.1% 1.1%
---[NoImage] x264 [info]: ref P L0 72.5% 16.0% 11.5%
---[NoImage] x264 [info]: ref B L0 80.8% 19.2%
---[NoImage] x264 [info]: ref B L1 95.9% 4.1%
---[NoImage] x264 [info]: kb/s:1997.0
---[NoImage] encoded 10001 frames, 50.93 fps, 1997.15 kb/s

http://www.users.on.net/~audionut11/subme6.mkv

Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 1 --bitrate 2000 --stats "F:\subme7.stats" --level 4.1 --keyint 240 --min-keyint 24 --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --subme 2 --partitions none --me dia --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output NUL "F:\test.avs"
--[Information] [4/07/2009 8:15:51 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile Main, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:21.08 size: 56017
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.20 size: 13289
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.85 size: 3966
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 46.3% 0.0% 53.7%
---[NoImage] x264 [info]: mb P I16..4: 17.3% 0.0% 0.0% P16..4: 50.2% 0.0% 0.0% 0.0% 0.0% skip:32.6%
---[NoImage] x264 [info]: mb B I16..4: 0.9% 0.0% 0.0% B16..8: 21.5% 0.0% 0.0% direct:12.0% skip:65.6% L0:26.4% L1:48.6% BI:25.0%
---[NoImage] x264 [info]: final ratefactor: 21.40
---[NoImage] x264 [info]: direct mvs spatial:98.4% temporal:1.6%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:33.9% 44.3% 12.2% inter:18.1% 12.0% 0.8%
---[NoImage] x264 [info]: kb/s:1926.3
---[NoImage] encoded 10001 frames, 95.20 fps, 1926.42 kb/s

Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 2 --bitrate 2000 --stats "F:\subme7.stats" --me hex --subme 7 --level 4.1 --keyint 240 --min-keyint 24 --ref 3 --mixed-refs --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output "F:\subme7.mkv" "F:\test.avs"
--[Information] [4/07/2009 8:17:37 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile High, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:20.67 size: 59468
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.06 size: 13489
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.33 size: 4440
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 11.9% 69.7% 18.4%
---[NoImage] x264 [info]: mb P I16..4: 1.2% 3.6% 0.6% P16..4: 46.0% 11.3% 11.6% 0.0% 0.0% skip:25.8%
---[NoImage] x264 [info]: mb B I16..4: 0.0% 0.3% 0.1% B16..8: 49.4% 1.0% 2.1% direct: 2.9% skip:44.1% L0:37.9% L1:55.7% BI: 6.4%
---[NoImage] x264 [info]: 8x8 transform intra:67.7% inter:55.2%
---[NoImage] x264 [info]: direct mvs spatial:87.0% temporal:13.0%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:66.4% 71.2% 37.5% inter:16.3% 15.7% 1.0%
---[NoImage] x264 [info]: ref P L0 72.7% 15.7% 11.5%
---[NoImage] x264 [info]: ref B L0 83.5% 16.5%
---[NoImage] x264 [info]: ref B L1 96.2% 3.8%
---[NoImage] x264 [info]: kb/s:1996.7
---[NoImage] encoded 10001 frames, 42.70 fps, 1996.80 kb/s

http://www.users.on.net/~audionut11/subme7.mkv

Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 1 --bitrate 2000 --stats "F:\umhme6.stats" --level 4.1 --keyint 240 --min-keyint 24 --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --subme 2 --partitions none --me dia --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output NUL "F:\test.avs"
--[Information] [4/07/2009 8:23:10 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile Main, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:21.08 size: 56017
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.20 size: 13289
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.85 size: 3966
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 46.3% 0.0% 53.7%
---[NoImage] x264 [info]: mb P I16..4: 17.3% 0.0% 0.0% P16..4: 50.2% 0.0% 0.0% 0.0% 0.0% skip:32.6%
---[NoImage] x264 [info]: mb B I16..4: 0.9% 0.0% 0.0% B16..8: 21.5% 0.0% 0.0% direct:12.0% skip:65.6% L0:26.4% L1:48.6% BI:25.0%
---[NoImage] x264 [info]: final ratefactor: 21.40
---[NoImage] x264 [info]: direct mvs spatial:98.4% temporal:1.6%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:33.9% 44.3% 12.2% inter:18.1% 12.0% 0.8%
---[NoImage] x264 [info]: kb/s:1926.3
---[NoImage] encoded 10001 frames, 94.80 fps, 1926.42 kb/s

Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 2 --bitrate 2000 --stats "F:\umhme6.stats" --me umh --subme 6 --level 4.1 --keyint 240 --min-keyint 24 --ref 3 --mixed-refs --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output "F:\umhme6.mkv" "F:\test.avs"
--[Information] [4/07/2009 8:24:56 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile High, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:20.62 size: 59994
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.00 size: 13601
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.29 size: 4274
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 11.8% 69.9% 18.3%
---[NoImage] x264 [info]: mb P I16..4: 1.0% 3.2% 0.6% P16..4: 45.9% 11.7% 12.4% 0.0% 0.0% skip:25.2%
---[NoImage] x264 [info]: mb B I16..4: 1.3% 0.0% 0.0% B16..8: 24.2% 1.4% 1.2% direct:10.7% skip:61.2% L0:32.4% L1:45.4% BI:22.2%
---[NoImage] x264 [info]: 8x8 transform intra:60.1% inter:51.5%
---[NoImage] x264 [info]: direct mvs spatial:87.0% temporal:13.0%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:61.6% 69.5% 35.5% inter:16.3% 18.2% 1.1%
---[NoImage] x264 [info]: ref P L0 72.1% 16.2% 11.7%
---[NoImage] x264 [info]: ref B L0 80.6% 19.4%
---[NoImage] x264 [info]: ref B L1 95.9% 4.1%
---[NoImage] x264 [info]: kb/s:1997.1
---[NoImage] encoded 10001 frames, 41.58 fps, 1997.18 kb/s

http://www.users.on.net/~audionut11/umhme6.mkv

Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 1 --bitrate 2000 --stats "F:\umhme7.stats" --level 4.1 --keyint 240 --min-keyint 24 --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --subme 2 --partitions none --me dia --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output NUL "F:\test.avs"
--[Information] [4/07/2009 8:28:59 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile Main, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:21.08 size: 56017
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.20 size: 13289
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.85 size: 3966
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 46.3% 0.0% 53.7%
---[NoImage] x264 [info]: mb P I16..4: 17.3% 0.0% 0.0% P16..4: 50.2% 0.0% 0.0% 0.0% 0.0% skip:32.6%
---[NoImage] x264 [info]: mb B I16..4: 0.9% 0.0% 0.0% B16..8: 21.5% 0.0% 0.0% direct:12.0% skip:65.6% L0:26.4% L1:48.6% BI:25.0%
---[NoImage] x264 [info]: final ratefactor: 21.40
---[NoImage] x264 [info]: direct mvs spatial:98.4% temporal:1.6%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:33.9% 44.3% 12.2% inter:18.1% 12.0% 0.8%
---[NoImage] x264 [info]: kb/s:1926.3
---[NoImage] encoded 10001 frames, 93.61 fps, 1926.42 kb/s

Job commandline: "C:\Users\Audionut\Desktop\x264_x86_r1173_techouse\x264.exe" --pass 2 --bitrate 2000 --stats "F:\umhme7.stats" --me umh --subme 7 --level 4.1 --keyint 240 --min-keyint 24 --ref 3 --mixed-refs --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --threads auto --thread-input --aq-mode 0 --sar 1:1 --progress --no-psnr --no-ssim --output "F:\umhme7.mkv" "F:\test.avs"
--[Information] [4/07/2009 8:30:47 PM] Encoding started
--[NoImage] Standard output stream
--[NoImage] Standard error stream
---[NoImage] avis [info]: 1280x544 @ 23.98 fps (10001 frames)
---[NoImage] x264 [info]: using SAR=1/1
---[NoImage] x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 Cache64
---[NoImage] x264 [info]: profile High, level 4.1
---[NoImage]
---[NoImage] x264 [info]: slice I:159 Avg QP:20.66 size: 59576
---[NoImage] x264 [info]: slice P:5631 Avg QP:23.05 size: 13479
---[NoImage] x264 [info]: slice B:4211 Avg QP:24.33 size: 4450
---[NoImage] x264 [info]: consecutive B-frames: 26.9% 45.6% 7.9% 19.6%
---[NoImage] x264 [info]: mb I I16..4: 11.8% 69.8% 18.3%
---[NoImage] x264 [info]: mb P I16..4: 1.0% 3.2% 0.5% P16..4: 46.1% 11.7% 12.1% 0.0% 0.0% skip:25.4%
---[NoImage] x264 [info]: mb B I16..4: 0.0% 0.2% 0.1% B16..8: 49.7% 1.0% 2.2% direct: 2.9% skip:43.9% L0:37.9% L1:55.8% BI: 6.3%
---[NoImage] x264 [info]: 8x8 transform intra:68.1% inter:55.0%
---[NoImage] x264 [info]: direct mvs spatial:87.0% temporal:13.0%
---[NoImage] x264 [info]: coded y,uvDC,uvAC intra:67.8% 72.1% 38.6% inter:16.0% 15.8% 1.0%
---[NoImage] x264 [info]: ref P L0 72.5% 15.9% 11.7%
---[NoImage] x264 [info]: ref B L0 83.4% 16.6%
---[NoImage] x264 [info]: ref B L1 96.2% 3.8%
---[NoImage] x264 [info]: kb/s:1996.7
---[NoImage] encoded 10001 frames, 35.42 fps, 1996.83 kb/s

http://www.users.on.net/~audionut11/umhme7.mkv

UMH too slow? Subme 7+HEX is slower than subme 6+UMH.

Not in my tests.

[kemuri-_9]

Not in my tests.

you tested umh+subme 6 against umh+subme7, not hex+subme7 as the claim condition states.
naturally umh+subme 7 would be slower than umh+subme 6....

[Audionut]

you tested umh+subme 6 against umh+subme7, not hex+subme7 as the claim condition states.
naturally umh+subme 7 would be slower than umh+subme 6....

Look again.

Tests are,

subme6+hex
subme7+hex
subme6+umh
subme7+umh

[kemuri-_9]

oh right, my miss, i was expecting --me hex and not the lack of a --me param (brain fart on hex being default temporarily)...

Edit:
knowing roozhou, he might throw out another condition to the statement....
how does testing on SD material look?

[Audionut]

hex being default

Thought about adding the param's there just for the sake of it.
Perhaps I should and use a little [b] on the important parts.

edit: wouldn't surprise me. FWIW, I choose 1280x544 as it's sort of good middle ground.
BTW, I'm also expecting being told off for not using enough b-frames, even though 3 could be classed as a default.

[destram]

Furthermore you should perhaps consider the fact that the speed penalty of umh compared to hex increases expotentially with soaring values in both subme and framerefs...
This has now nothing to do with the presets, of course, but I mostly experienced umh has a pretty lower profit than subme 6-->7

I tested various combinations of these factors to decide which one to use for encoding st:e for a friend...

I finally chose subme7 and hex, since umh didn't necessarily resulted in a better psnr/ssim, but slowed down encoding considerably.

What I still don't understand, is why an increase of subme usually soars the bitrate...
At least subme 6-->7 by around 1,8% and 7-->8 by about 1,0%

UMH and frameref respectively usually both let psnr/ssim rise and decrease the bitrate...

[roozhou]

@Audionut
You should test with DS's new default preset except me and subme. Different bitrates will affect frame-type decision, resulting in different speed on subme 6.

[10L23r]

exclude settings or setting combinations which may bring both worse quality and slower encoding speed.

hypothetically... what if i have a sample for which me hex is worse than me dia. or trellis 1 is worse than no trellis. or 5 bframes is worse than no bframes. i'm pretty sure that many options that usually help quality at the expense of some speed COULD potentially hurt quality in some samples. but, the chances of encountering such a sample is so low that one can justify using whatever option helps quality in the majority of samples, even though that setting may hurt quality is a minority of samples.

so if some setting "may bring worse quality and slower encoding speed," that doesn't mean that it should be excluded completely.

if you don't know already, subme 6 never gives the best quality of the three options: subme 5, 6, and 7.

[Audionut]

@Audionut
You should test with DS's new default preset except me and subme.

No, you should. The conditions were subme6+umh is faster than subme7+hex and quality of subme7 is worse that subme6.

Different bitrates will affect frame-type decision, resulting in different speed on subme 6.

--pass 2 --bitrate 2000 --stats "F:\subme6.stats"
encoded 10001 frames, 50.93 fps, 1997.15 kb/s

--pass 2 --bitrate 2000 --stats "F:\subme7.stats"
encoded 10001 frames, 42.70 fps, 1996.80 kb/s

--pass 2 --bitrate 2000 --stats "F:\umhme6.stats"
encoded 10001 frames, 41.58 fps, 1997.18 kb/s

--pass 2 --bitrate 2000 --stats "F:\umhme7.stats"
encoded 10001 frames, 35.42 fps, 1996.83 kb/s

[10L23r]

What I still don't understand, is why an increase of subme usually soars the bitrate...
At least subme 6-->7 by around 1,8% and 7-->8 by about 1,0%


i think it's cus the increase in subme requires more psy-rd, and since psy-rd lowers metrics, x264 is fooled into thinking that more bitrate is required, even though the opposite is true in most cases. just do a quick comparison of subme 7 w/o psyrd at crf 22 and subme 7 with psyrd at crf 22. the latter will be significantly larger.

[roozhou]

No, you should. The conditions were subme6+umh is faster than subme7+hex and quality of subme7 is worse that subme6.

DS's default uses crf 23. Why not run 1st pass with crf 23 instead of bitrate 2000?

[Selur]

DS's default uses crf 23. Why not run 1st pass with crf 23 instead of bitrate 2000?
depending on your source a bitrate of 2000 could result in an average quantizer that could be far away from crf 23. So why do you want to use crf X if you aim for a specific bitrate filesize?

[Audionut]

with crf 23 instead

I clearly recall the reasons why being posted in other threads you have participated in. :search:

[10L23r]

DS's default uses crf 23. Why not run 1st pass with crf 23 instead of bitrate 2000?


It's also possible to run the first pass of a 2-Pass encode with CRF mode (instead of ABR mode), but the second pass will be ABR again (with the target bitrate that resulted from the first CBR pass). But I say it again: This kind of "2-Pass CRF" makes only sense, if you want to use CRF mode with VBV (because VBV doesn't perform well with a single pass). However if you are not using VBV, then CRF will work perfectly fine with a single pass! And if you want to do a "classical" 2-Pass encode to hit a desired filesize (or target betirate), then you can not run the first pass in CRF mode for obvious reasons...


@roozhou: so what are you suggesting for defaults? i'm kinda lost... do u want subme7+umh or subme6+umh? r u saying that subme6+umh is better (in terms of quality/speed efficiency) than subme7+hex?

[roozhou]

@roozhou: so what are you suggesting for defaults? i'm kinda lost... do u want subme7+umh or subme6+umh? r u saying that subme6+umh is better (in terms of quality/speed efficiency) than subme7+hex?

I changed my mind.
Now I suggest keeping subme6 + hex as default.