dvd-rb, cce and multiple cores

[drmih]

Having been playing with bd-rb and seeing the way that the programme, or rather x264, can utilise all of the cores of an i7, I have been trying to do a dvd. dvd-rb has multithreading set, and identifies (8), which I'm guessing is the number of threads. However, when encoding a dvd cce is only using about a third of the capability (typically 35% of the cpu, with the remainder being 'idle') - it maxes out 1 core and half of the next. It's still flying even with the maximum number of passes but I wondered if there was a reason for not grabbing the remaining 5 threads?

[jdobbs]

It depends which encoder you are using. I believe CCE has a limited number of threads. But with HC you should be running 8 instances of the program -- and I would think that uses a lot more than 35% (and a lot of memory too, though).

[drmih]

Thanks - I've never used anything but CCE. I'll give HC a go. In your opinion does it compare favourably quality wise?

**Just trying - HC is using all 8 threads by launching 8 individual process on 8 segments - typically 99% utilisation and nearly 4Gb memory

*** The problem with HC seems to be that it uses one thread per segment. Although it starts off with 8 threads, it quickly gets through the 'small' ones and then you end up with less and less threads in bigger segments. I'm already down to 4 now. Maybe there is a way for it to use both threads of a core on a single segment. *****Just did another disc and the segments were better balanced and there were many more process running for the majority of the time.

**** Did a dvd in the highest quality in around 20 mins, so impressive but I don't know how that compares with CCE 9 pass which was around 3 hours

[turbojet]

HCenc is a great free alternative to CCE. I've done some tests comparing the 2 visually. First I viewed them individually and I couldn't choose one over the other however after that I compared them to source and CCE looked more like the original. The main difference was HC had a slightly redder tint then the source and cce. Beyond that I think cce retained grain a little better but I don't think you'd notice this at 24/25/30 fps. I used HCenc 0.23 and CCE SP2.

Unless you are encoding often with low bitrates (<3000 for a typical source) or having sizing issues with 4-5 cce passes I can't see any reason using more CCE passes.

[jdobbs]

Anything over 3 pass in CCE has negligible quality impact. I never go over three. Custom Technologies has said it hits the max at about 4 passes in previous CCE documentation. The other passes are just there to make us all feel good.