Upcoming version - Performance boost

[Pete]

For those who have been unhappy with conversion speeds....

I have just had a little playtime with a pre-beta.
I'm happy to say, that you will be pleased. Transcoding speeds have at least doubled, some videos are processed at even three times the speed previously observed.

The guys have been mainly concentrating on performance optimization for the upcoming beta.

 

[Ch3vr0n]

As long as the video quality doesn't suffer negatively more speed is always good

Verstuurd vanaf mijn Nexus 7 met Tapatalk

 

[whatever_gong82]

Overall, I love all the new changes and improvements in CloneBD. It is way better than when it came out back last December.

But, to re-iterate what Ch3vr0n said, as long as the video quality doesn't suffer, I'm all for it.

 

[Pete]

Before speculation goes on: the speed improvements have nothing to do with quality.
No changes have been made to codec settings.

 

[Ch3vr0n]

I want speculating, but in that case that is indeed very good news. Speed is always welcome.

Verstuurd vanaf mijn Nexus 7 met Tapatalk

 

[raybestos]

I am glad hear of a performance boost. More details will be appreciated. Will it be to be libav core, or the 3D CUDA core, will it be both for the 32-bit libav and 64-bit libav core? 2-3x seems like something must give. Either that, or the existing compiles of libav were optimized very poorly, indeed!

 

[gereral1]

Just curious.... Speed can be affected based on a single or dual pass when encoding. At least this is what I've experienced using h264 codec. Is this how they are doubling the speed? I'm not a user of clonebd but have been watching this closely.

 

[Pete]

I am glad hear of a performance boost. More details will be appreciated. Will it be to be libav core, or the 3D CUDA core, will it be both for the 32-bit libav and 64-bit libav core? 2-3x seems like something must give. Either that, or the existing compiles of libav were optimized very poorly, indeed!

No CUDA. yet, no.
A couple of bottle necks were identified and trashed. And yes, the libav optimization was not ideal due to bugs in gcc optimization (there was a choice of either slow processing or crashes). An updated version of gcc fixed that problem.
Also parallelization was optimized, some directx-shortcomings worked around and several non-libav routines rewitten with SIMD instructions.

Just curious.... Speed can be affected based on a single or dual pass when encoding. At least this is what I've experienced using h264 codec. Is this how they are doubling the speed? I'm not a user of clonebd but have been watching this closely.

Dual pass is just what it says: "do it twice". And CloneBD doesn't do it, has never. Dual pass simply transcodes the same input twice.
First run is meant to gather bitrate information, second run then transcodes and boosts the bit rate on the more complex scenes at cost of the less complex ones.

That feature is probably one of the most misused ones (and least understood).
It is useful when you're compressing a video down to a really bad quality (e.g. when you're planning to stream it over a slow network or simply have no space to waste).
Quality can improve significantly - by making those demanding, complex scenes more bearable (you may have seen streaming video in the early days, heavily pixelated and the picture tended to pretty much break down, when there was more movement in the scene).

But when used on typical use cases for CloneBD, the difference is hardly percievable and an enormous waste of time IMO.

 

[gereral1]

No CUDA. yet, no.
A couple of bottle necks were identified and trashed. And yes, the libav optimization was not ideal due to bugs in gcc optimization (there was a choice of either slow processing or crashes). An updated version of gcc fixed that problem.
Also parallelization was optimized, some directx-shortcomings worked around and several non-libav routines rewitten with SIMD instructions.



Dual pass is just what it says: "do it twice". And CloneBD doesn't do it, has never. Dual pass simply transcodes the same input twice.
First run is meant to gather bitrate information, second run then transcodes and boosts the bit rate on the more complex scenes at cost of the less complex ones.

That feature is probably one of the most misused ones (and least understood).
It is useful when you're compressing a video down to a really bad quality (e.g. when you're planning to stream it over a slow network or simply have no space to waste).
Quality can improve significantly - by making those demanding, complex scenes more bearable (you may have seen streaming video in the early days, heavily pixelated and the picture tended to pretty much break down, when there was more movement in the scene).

But when used on typical use cases for CloneBD, the difference is hardly percievable and an enormous waste of time IMO.

Thank-you for explaining dual pass. I always thought the second pass was for frame error checking. I always use dual pass every movie and it takes 4-5 hours to compress at 15,000 bit rate. I can usually take a 24 gig down to 12-14gig. I usually do this before I goto bed. Anyway my picture quality is excellent. I learned something new today. Thks pete

 

[Pete]

Thank-you for explaining dual pass. I always thought the second pass was for frame error checking. I always use dual pass every movie and it takes 4-5 hours to compress at 15,000 bit rate. I can usually take a 24 gig down to 12-14gig. I usually do this before I goto bed. Anyway my picture quality is excellent. I learned something new today. Thks pete

Maybe you can convince me, that dual pass is worth it when you compare it to a single pass result
But 15000 is ample and I'd expect very good quality either way.

 

[Soyado]

For those who have been unhappy with conversion speeds....

I have just had a little playtime with a pre-beta.
I'm happy to say, that you will be pleased. Transcoding speeds have at least doubled, some videos are processed at even three times the speed previously observed.

The guys have been mainly concentrating on performance optimization for the upcoming beta.

Hi Pete,

Thanks......more speed will be a great improvement. There is a problem with keeping the movie menu in partial backup. Some types of menus cannot be kept and functional. Do you know if this issue will be fixed soon ?

Thanks again

 

[Pete]

Hi Pete,

Thanks......more speed will be a great improvement. There is a problem with keeping the movie menu in partial backup. Some types of menus cannot be kept and functional. Do you know if this issue will be fixed soon ?

Sorry, presently not on the immediate todo list. Some other things are more urgent and need to be released.

 

[ddjmagic]

I have just had a little playtime with a pre-beta.
I'm happy to say, that you will be pleased. Transcoding speeds have at least doubled, some videos are processed at even three times the speed previously observed.

Sounds very promising, looking forward to trying it

 

[teragor]

For those who have been unhappy with conversion speeds....

I have just had a little playtime with a pre-beta.
I'm happy to say, that you will be pleased. Transcoding speeds have at least doubled, some videos are processed at even three times the speed previously observed.

The guys have been mainly concentrating on performance optimization for the upcoming beta.

along with speed does it improve mkv /mp4/ subtittles?

 

[teragor]

As long as the video quality doesn't suffer negatively more speed is always good

Verstuurd vanaf mijn Nexus 7 met Tapatalk

ch3vr0n do you know if it will improve across the board... things like mkv mp4 subs also ?????

 

[Ch3vr0n]

Nope, i dont work for slysoft. But it should. encoding engine improvements affect everything, its only the output that's different, its still the same engine being used.

 

[Gummigutta]

Nope, i dont work for slysoft. But it should. encoding engine improvements affect everything, its only the output that's different, its still the same engine being used.

So the output will be different compared to prior versions - even if the company behind the program stated before that it shouldn't be?

 

[Ch3vr0n]

I meant different as in eg 'BD25 vs mkv'. Full disc clone vs portable file.

Verstuurd vanaf mijn Nexus 7 met Tapatalk

 

[gereral1]

I am wondering about keeping menu option. Most users would want movie only with audio and subtitle options in menu?

Maybe you can convince me, that dual pass is worth it when you compare it to a single pass result
But 15000 is ample and I'd expect very good quality either way.

Maybe a lab test showing 2 videos playing side by side would be worth a try. If I have time ill compress a action part and slow action part and play them side by side and see if I can see any difference with 1 or 2 passes.

 

[Ch3vr0n]

most users want movie only, i doubt that. I think most users want a full disc copy or full disc without the trailer crap. I'm one of those