multi-cpu in multi-encoder mode

[dazednconfuzed]

I was wonder in multi encoder "rip to" mode does multi-cpu work. i'm using wma and frauhofer mp3.the bios is setup to be duo core enabled. It "ripping" the cd and then cpu1 encoding.

any suggestion would be appreciated.

[MikeCerm]

Re: multi-cpu in multi-encoder mode

Supposedly, multi-CPU support is a feature limited to the "Reference" version. However, it works fine with my dual-core system (Powerpack, registered). It does depend on what formats you're using. Certain codecs are inherently multithreaded (WMA being one of them), and some are not (like LAME, for example). I certainly don't have any problems with the multi-encoder using both of my cpu cores.

[dazednconfuzed]

Re: multi-cpu in multi-encoder mode

I am using reference. Is there any mp3 encoder that's multithreaded.

[Spoon]

Re: multi-cpu in multi-encoder mode

CPU 1 encoding is only the message that cd ripper knows about, multi encoder cannot pass back to cd ripper, open task manager when encoding and look a cpu usage.

[MikeCerm]

Re: multi-cpu in multi-encoder mode

I am using reference. Is there any mp3 encoder that's multithreaded.

Yes. LAME MT is a project aimed at creating a multithreaded version of LAME. You can use it with the CLI encoder.

http://softlab.technion.ac.il/project/LAME/html/lame.html

http://www.lame-mt.com/

Recommended settings: -V 2 --vbr-new -q1 --mt - [outfile]
(the number after the 'V' determines the bitrate (lower number = higher bitrate), and the number after 'q' affects quality/speed (lower number is means slower, but higher quality encoding).

On a dual-core system, the speed increase is about 30-40%. However, there is a catch: it's built on version 3.97a or 3.97b2 (vs. 3.97 stable, natively in dMC). You won't hear a difference, and it won't crash or anything, but you could make an esoteric case that the multithreaded version somehow lower quality, or less stable.

If you want the best quality regardless of speed, don't use the MT version. If you're willing to give up some quality for speed, MT is great.

[bhoar]

Re: multi-cpu in multi-encoder mode

You won't hear a difference, and it won't crash or anything, but you could make an esoteric case that the multithreaded version somehow lower quality, or less stable.

If you want the best quality regardless of speed, don't use the MT version. If you're willing to give up some quality for speed, MT is great.

Last I read, Lame MT doesn't use the bit reservoir, so you're sacrificing some audio quality at the same bitrates...

-brendan

[Spoon]

Re: multi-cpu in multi-encoder mode

See this test using Lame MT on a quad system:

http://www.techreport.com/reviews/2007q2/pricevperf/index.x?pg=6

It is not that impressive, when considering if dBpoweramp Reference was used to encode 100 files to lame, it would be 4x faster than the single core equivalent.

[MikeCerm]

Re: multi-cpu in multi-encoder mode

Last I read, Lame MT doesn't use the bit reservoir, so you're sacrificing some audio quality at the same bitrates...

-brendan

In CBR mode, where bits are used inefficiently in many frames, it is useful to save those bits for more complex frames. It is my understanding that, when using VBR, the bit reservoir doesn't really make that much of a difference, because the bitrate is able to change to accommodate the complexity. Is there a difference? Yes, but it's certainly not significant to the average person. If you believe that you can hear a difference between a VBR file with and without the bit reservoir, you really shouldn't be using MP3 anyway. Ogg, WMA, and AAC will all produce far better results (as will lossless), and one of those formats is bound to play on whatever device you listen to music on.

It is not that impressive, when considering if dBpoweramp Reference was used to encode 100 files to lame, it would be 4x faster than the single core equivalent.

The 30-40% increase that LAME MT provides (on dual-core systems) is pretty impressive. However, you're totally right: When it comes to LAME, using dMC Reference to run separate conversion processes per core in parallel will produce better, faster results when converting more than one file. On the other hand, when using WMA, which is much better at fully utilizing multiple cores (~80% speed increase per core), running multiple instances simultaneously would not have quite the same benefit (compared to running LAME in parallel).

[foylefamily]

Re: multi-cpu in multi-encoder mode

Thought I'd post my question here before I email support...

I have Reference and it does support my dual processors, but I have noticed that when it is processing 17,000 files conversions that it will get so far then drop one of the processors, and so slow things down. Is there a bug, or if there a condition that can cause Reference to shut down using one of the processors?

[LtData]

Re: multi-cpu in multi-encoder mode

What formats are you converting to? I personally have converted my ~2,000 file collection and it always uses both CPUs. If you do the converting in smaller batches, say ~5,000, does one core quit working then too?

[foylefamily]

Re: multi-cpu in multi-encoder mode

What formats are you converting to? I personally have converted my ~2,000 file collection and it always uses both CPUs. If you do the converting in smaller batches, say ~5,000, does one core quit working then too?

Yes. It doesn't really matter the size of the batch, I even dropped it down to alphabetical Artist in the selections, and even with a few hundred conversions it still dropped CPU 2 at some point. I only notice because when I go back after a period of time, I see CPU2 doing nothing, but CPU1 is doing all the work, and the time estimate for completion has increased significantly (which is fairly obvious when only one CPU is working).

If I stop the process and start it again, then 2 CPUs are in use again for awhile.

I wonder if there could be an outside influence, or perhaps Reference checks the load on the CPU, and drops it, but never picks it up again... I haven't got a clue...

[foylefamily]

Re: multi-cpu in multi-encoder mode

What formats are you converting to?

I forgot to mention, that I was converting from MP3 to WMA.

[Spoon]

Re: multi-cpu in multi-encoder mode

Perhaps one of the encoders has stalled, perhaps a bad file.

[foylefamily]

Re: multi-cpu in multi-encoder mode

Perhaps one of the encoders has stalled, perhaps a bad file.

Do you mean bad input file (one of the MP3s was bad), or the codec has a bad file?

Would it stop the processor, or shouldn't it just report the error and continue on? I guess I am still a little confused.

[LtData]

Re: multi-cpu in multi-encoder mode

I think Spoon is referring to one of your mp3 files being bad.