|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
198.54.202.195
In Reply to: I'm not sure I understand you posted by Christine Tham on April 7, 2007 at 19:37:45:
Of all latencies in computer, I find reducing playback latency to have a large impact on sound. Sound delivery happens in realtime, hence preference for soundcard having lions share of PCI bus.On SATA/IDE, SATA latency may be worse than IDE (but biggest latency is at hard drive irrespective of interface @ ~5-6ms). Of course, you raise a valid point why this latency is less important?
My thinking: disk read latency happens only once at start of play where Foobar through Windows (through interface, data bus, disk controller, cache and finally disk) establishes a data read stream. Since this sequential read stream rate (@ 42.9MB/s for Seagate 5400.3 160GB) is factors more than what Foobar needs there is minimal effect (as disk data is available to Foobar ahead of time –Windows/disk controller prefetch & block reads occur). For example, on my setup, an hour long CD (600MB) takes ~10 seconds to load into RAM, yet playback takes an hour.
Let me preempt your question here, 600/42.9 = 14 seconds, how do you get to ~10 seconds? RAID 0. File data is stripped sequentially across drives thus on read, you get both drives performing sequential read in parallel. In theory you should get double throughput but some serialization occurs and throughput improves by ~30-50%. RAID 0 does away with Windows IDE drivers (and all its overheads including SATA in IDE mode) which I found beneficial.Most optimal is when NO disk reading occurs during playback. By this I mean entire CD .wav file is read into RAM (using Foobar’s full file buffing). There is a bug when playing .cue files where foobar ignores file buffering – playing .wav files, buffering occurs (and 5-15 seconds elapses for playback to start). Hopefully this bug gets fixed in next release. This way, disks go into idle state consuming 40% less power (meaning less EMI, etc.).
*** If the DAC is reclocking and resamping the SPDIF input, then why optimize the PC at all (since any jitter in the input signal to the AA will be discarded by it's resampling)? ***There is good sonic improvement. Like all DACs I suppose the one I’m using is not perfect. Others have found this too with DACs that do internal jitter management.
*** Also, why upsample in Foobar using SRC? ***
To bypass DAC’s inferior 16/44.1 -> 24/96 upsampling. DAC further upsamples to 24/192 but most ‘damage’ occurs on former. DAC sees 24/96 feed as native input (with music information to 22.05kHz only).
Follow Ups:
Two immediate points:1. Raid - unless you have a premium raid card, raid support is software based.
Recommended Windows settings as set out achieves this by:1. Having no paging file (no VM)
2. Stopping all unnecessary services (task scheduler, disk indexing, etc.)
3. Remove unnecessary programs (many sw installations do sneek in startup actions)
4. Remove Windows components (disk indexing, messenger, etc.)
5. Only have foobar in startup
6. Perform boot only prefetch (and deleting all contents of c:\windows\prefetch)
7. Install no virus protection swTest this and let me know what you get. After bootup, with Foobar started but not playing, there should be no disk activity whatsoever no matter how long system idles. Only disk activity would be the disks themselves (Windows has no control over this), e.g. parking drive heads and thermal recalibration.
Recommended Biostar mobo implements SATA RAID at hardware level. Check out the VIA chipset it uses (link below). No need to get additional RAID hardware if using this mobo.Once BIOS is set to SATA RAID, Windows installation CD will not recognize disk. VIA chipset driver for Windows needs to be supplied on floppy.
| ||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: