Poor Performance using Batch Converter (dBPoweramp 14.3)

[doug.brown]

I've been using batch converter for several years to convert large numbers (~20,000) of FLAC files to Nero AAC. With my older Nehalem (first generation Core i7) based machine with 8 GB RAM and WD Cavier Black HDDs, I would get encoding speeds in the range of 160X. Last summer, I moved to a new Ivy Bridge (thrid generation Core i7) machine with 32GB RAM and fast WD Velociraptor HDDs. Encode speed increase to around 280X. During encoding, Task Manager revealed the machine to be CPU bound, as expected with 8 instances of neroaccenc.exe consuming almost all CPU resources. Lately, however, using exactly the same settings I am only getting encode speeds in the 60X range. Task Manager shows the machine to be HDD bound, with CPU only running ~20-25%. I have 8 HDDs in the PC so I tried using different combinations of HDD for the source and destination files and it made no significant difference. Any ideas of what else to check?

[doug.brown]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

BTW, I am using Windows 8, 64 bit and have been since building the new machine.

[Spoon]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

as a test try a low bit rate (should have less hdd usage), what is the encode speed?

[doug.brown]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

The target bit rate doesn't seem to make much difference. I tried Nero ACC at q0.25, q0.55 and q0.95 and observed encoding at around 60X for each. It is the HDD with the source FLAC files that hits 100% utilization. I've moved the source FLAC files to different drives and it make no difference to encode speed. The HDDs are defragmented. Could it be an issue with the FLAC decoder? I do seem to get long periods of "Rendering for DSP effects" even though I only have the remove silence effect selected. I'll try removing all the DSP effects and see if it makes a difference.

[dbfan]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

It will be the dap effect as the audio is decoded and written to tmp folder.

[doug.brown]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

It will be the dap effect as the audio is decoded and written to tmp folder.

How can I figure out out where it is writing the temp files? I suspect it may be writing to the C: drive which is an SSD and has slow write performance. I manually remapped many of the temp file settings away from the C: drive and may have missed some. The C: drive does not show high utilization under Task Manager.

[Spoon]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

if you type into explorer %tmp% this is where it writes.

[doug.brown]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

Discovered a few things:
-%tmp% was directed to the D: HDD (not to the C: SSD), as expected
-removing the ReplayGain DSP effect makes no difference
-removing the Trim Silence effect seems to resolve the issue and allows the encode speed to go to ~270X
-with Trim Silence enabled, the utilization of the HDD containing the source FLAC (F files goes to 100%, utilization of the HDD with %tmp% (D goes up, but is still below 25%, utilization of the destination drive (E for the AAC files is around 30%

Why would utilization of the source drive red line given that DBPoweramp has to read the files for conversion regardless so trimming silence shouldn't add load to it? Could there be something wrong with the FLAC decoder or the Trim Silence DSP effect? I'm sure I had Remove Silence enabled in the past and was able to get 270X encoding speeds.

Could I run all the FLAC files though a FLAC to FLAC conversion with Remove Silence enabled first and then never have to include Remove Silence when doing batch conversion to AAC? I could probably do the same with Replay Gain.

A bit of background...I have around 20,000 FLAC files stored on a NAS that I stream to several clients throughout the house. I'm constantly adding to the FLAC library. About once a month, I copy all of the FLAC files to an HDD on a desktop PC. Using the desktop PC, I do three separate batch conversions:
1) Target 350 kbs ACC using neroaacenc to copy to my car which has 256GB of storage
2) Target 192 kbs AAC using neroaacenc to copy to my phone which has 128GB of storage
3) Target 160 kbs AAC using neroaacenc to copy to my iPod Nano which has 16 GB of storage (have to be selective as not enough storage for all 20,000 songs)

[Spoon]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

What is drive F?

[doug.brown]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

Not sure why drives turned into emoticons. Setup is:

Drive C: SSD - not used at all by DBPowerAmp
Drive D: very fast WD Velociraptor HDD - temp files
Drive E: WD Caviar Black HDD - destination for encoded AAC files
Drive F: WD Caviar Black HDD - source for FLAC files to be encoded

[Spoon]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

What MB/s values are you getting for the drives at 100%?

[doug.brown]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

Encoding FLAC files from F: to Q0.535 Nero ACC on E: with ReplayGain and Trim Silence effects:
Overall ~70X encoding, CPU averaging ~20%
D: highest read ~42 MB/s, hihgest write ~70 MB/s
E: highest read ~7 MB/s, highest write ~18 MB/s (why are there reads?)
F: highest read ~12 MB/s, write is zero

Encoding FLAC files from F: to Q0.535 Nero ACC on E: with only ReplayGain effect:
Overall ~140X encoding, CPU averaging ~50%
D: highest read ~1 MB/s, highest write ~5 MB/s
E: highest read ~9 MB/s, highest write ~16 MB/s (why are there reads?)
F: highest read ~18 MB/s, write is zero

Encoding FLAC files from F: to Q0.535 Nero ACC on E: with no DSP effects:
same as with ReplayGain effect


I think something is up with drive E

[Spoon]

Re: Poor Performance using Batch Converter (dBPoweramp 14.3)

You said earlier that F had 100%, for reading 20MB/s it should not be 100%

You can have reads on the writing drive, because the file has to be read to write ID tags after encoding.