[paranoia] dae speed on plextor cdrw

[Chad Carr]

On Sat, Feb 16, 2002 at 02:28:33AM +0200, Sami Farin wrote:
> On Thu, Feb 14, 2002 at 09:08:00PM -0800, Chad Carr wrote:
> > On Fri, Feb 15, 2002 at 02:32:13AM +0200, Sami Farin wrote:
> > > On Fri, Feb 15, 2002 at 12:27:34PM +1300, John Morton wrote:
> > > ...
> > > > That said, it's not necessarily the case that the patch will actually speed 
> > > > up extraction for every drive and system out there, and cdparanoia does it's 
> > > > magic by doing multiple reads, which will always slow down the proceedings. 
> > > 
> > > With PIO, there's some maximum speed limit - for me it's 4 MB/s
> > > (and 12 MB/s with UDMA33) when reading from ATA-4 hard disk.
> > > So you might not reach ATAPI device's maximum DAE extraction speed
> > > when using PIO. 
> > 
> > Okay, patched and booted.  Ripping occurred.  Results:
> > 
> > LiteOn LTD163: _exactly_ the same
> > Plextor PX-W2410A: nearly _twice_ as slow (64 min cd in 9:50)
> > 
> > What else could I be doing wrong?  Did I mention that I am using ide-scsi?
> > Maybe that has something to do with it?
> 
> No, I don't think ide-scsi is related to ripping speed.
> 
> Can you show "hdparm -i" output for your Plextor and LiteOn
> after ripping?
> And if you have chipset-specific file (like "piix") at /proc/ide,
> can you show it, too...

Thanks very much for your assistance.  Please find attached
hdparm_i_liteon.output, hdparm_i_plextor.output and proc_ide_via.output.  Both
of the cdroms are plugged into the via controller, hdc and hdd, respectively.
I am not using them to rip at the same time.

> Have you seen messages "DMA disabled" or anything now that you
> use the ide-akpm patch?
> Do you get the "old speed" when you disable DMA with hdparm?

I never got those messages.  Where would I be looking for them?  In dmesg?

I do go right back to the "old speed" if I remove DMA.  Also, I am setting "-c1
-u1" in addition to "-d1" with hdparm.

After some more extensive testing, I realize that when I said that the Plextor
became "twice as slow" (I believe I even underlined it) this was not true.  It
is still slower but not as bad as I had originally reported.  With new and more
methodical testing, I got the following results (ranked highest to lowest) with
the Plextor on an audio CD with 326132 sectors (approx 72 min):

no DMA/no paranoia - 282s (15.42x)
with DMA/no paranoia - 323s (13.46x)
with DMA/with paranoia - 675s (6.44x)
no DMA/with paranoia - 676s (6.43x)

Also, just a question, is sectors per second deterministic if I don't use
paranoia?  Or can it vary based upon the CD I am using?  I know that a longer
CD is going to push up the average speed since the outside spins faster
(relative to the head) than the inside, but other than that, are there
differences between disks?

> And have you played with this jumper?
> http://www.plextor.com/english/news/dmajumper.html

I heard about this, but heard that DMA is enabled by default, and since the
kernel seems to recognize it as a DMA device on boot...okay I was lazy.  I'll
try it.

> I happened to see akpm on IRC and I asked about this Plextor "issue",
> he said "with IDE and CDROMs, anything is possible." :)