Mailing List Archive: mythmusic/cdparanoia question (a little OT).

mythmusic/cdparanoia question (a little OT).

Nov 28, 2002, 11:41 AM

Post #1 of 5 (1406 views)

Hi,

I'm working on the ripping/encoding at the same time (and also adding cddb
info in the tables as well as other music refinments) but I would like to
know if you too have a cpu usage of 60 to 90 % while ripping an audio cd.

My machine is a : Nforce/dual bank memory/AMD XP 1800+/Toshiba DVD/CD drive
DMA enabled for all drives, +- 20 MB/s transfer for disks

Also I have another question (maybe more to isaac).

I have more or less the same hardware setup as you have but can barely
record/play at 640/400, I tried all settings to no avail.

What kernel do you use, what tweaking can I do to reach your marks, does V42L
improve THAT much the performance ?

Christian

Re: mythmusic/cdparanoia question (a little OT). [ In reply to ]

mdz at debian

Nov 27, 2002, 5:10 PM

Post #2 of 5 (1400 views)

Permalink

On Thu, Nov 28, 2002 at 06:41:58PM +0000, Christian Morant wrote:

> I'm working on the ripping/encoding at the same time (and also adding cddb
> info in the tables as well as other music refinments) but I would like to
> know if you too have a cpu usage of 60 to 90 % while ripping an audio cd.

Welcome to the craptastic world of IDE. I have a 40x IDE CD-ROM and a 32X
SCSI CD-RW in this machine. While the IDE drive does CDDA slightly faster,
it uses about 1000% more CPU than the SCSI drive.

--
- mdz

Re: mythmusic/cdparanoia question (a little OT). [ In reply to ]

radford+myth at blackbean

Nov 28, 2002, 8:47 AM

Post #3 of 5 (1382 views)

Permalink

On Wed, Nov 27, 2002 at 07:10:20PM -0500, Matt Zimmerman wrote:
> On Thu, Nov 28, 2002 at 06:41:58PM +0000, Christian Morant wrote:
> > I'm working on the ripping/encoding at the same time (and also adding cddb
> > info in the tables as well as other music refinments) but I would like to
> > know if you too have a cpu usage of 60 to 90 % while ripping an audio cd.

> Welcome to the craptastic world of IDE. I have a 40x IDE CD-ROM and a 32X
> SCSI CD-RW in this machine. While the IDE drive does CDDA slightly faster,
> it uses about 1000% more CPU than the SCSI drive.

Nope, that's just linux that's crappy (inspired by the initially
crappy IDE cdroms) as it uses PIO instead of DMA when talking to the
cdrom. The newer 2.5 kernels don't have this problem thanks to Jens
Axboe.

Jens Axboe <axboe@suse.de> on Fri, 18 Oct 2002 writes:
> This patch should make 2.5 ide-cd writing even faster than using
> ide-scsi. Data is read/written directly to/from user space pages, and it
> is done using dma whenever possible.

-Jim

Re: mythmusic/cdparanoia question (a little OT). [ In reply to ]

mdz at debian

Nov 28, 2002, 9:07 AM

Post #4 of 5 (1396 views)

Permalink

On Thu, Nov 28, 2002 at 07:47:53AM -0800, Jim Radford wrote:

> On Wed, Nov 27, 2002 at 07:10:20PM -0500, Matt Zimmerman wrote:
> > On Thu, Nov 28, 2002 at 06:41:58PM +0000, Christian Morant wrote:
> > > I'm working on the ripping/encoding at the same time (and also adding cddb
> > > info in the tables as well as other music refinments) but I would like to
> > > know if you too have a cpu usage of 60 to 90 % while ripping an audio cd.
>
> > Welcome to the craptastic world of IDE. I have a 40x IDE CD-ROM and a 32X
> > SCSI CD-RW in this machine. While the IDE drive does CDDA slightly faster,
> > it uses about 1000% more CPU than the SCSI drive.
>
> Nope, that's just linux that's crappy (inspired by the initially
> crappy IDE cdroms) as it uses PIO instead of DMA when talking to the
> cdrom. The newer 2.5 kernels don't have this problem thanks to Jens
> Axboe.
>
> Jens Axboe <axboe@suse.de> on Fri, 18 Oct 2002 writes:
> > This patch should make 2.5 ide-cd writing even faster than using
> > ide-scsi. Data is read/written directly to/from user space pages, and it
> > is done using dma whenever possible.

Here's a link to the actual message and patch for anyone who's interested:

http://www.uwsg.iu.edu/hypermail/linux/kernel/0210.2/0935.html

This sounds great...but why do I see a similar (though less drastic)
difference between IDE and SCSI CPU utilization for hard disks, even when
DMA is enabled?

--
- mdz

Re: mythmusic/cdparanoia question (a little OT). [ In reply to ]

radford+myth at blackbean

Nov 28, 2002, 9:44 AM

Post #5 of 5 (1397 views)

Permalink

On Thu, Nov 28, 2002 at 11:07:05AM -0500, Matt Zimmerman wrote:
> On Thu, Nov 28, 2002 at 07:47:53AM -0800, Jim Radford wrote:
> > On Wed, Nov 27, 2002 at 07:10:20PM -0500, Matt Zimmerman wrote:
> > > Welcome to the craptastic world of IDE. I have a 40x IDE CD-ROM and a 32X
> > > SCSI CD-RW in this machine. While the IDE drive does CDDA slightly faster,
> > > it uses about 1000% more CPU than the SCSI drive.

> > Nope, that's just linux that's crappy (inspired by the initially
> > crappy IDE cdroms) as it uses PIO instead of DMA when talking to the
> > cdrom. The newer 2.5 kernels don't have this problem thanks to Jens
> > Axboe.

> This sounds great...but why do I see a similar (though less drastic)
> difference between IDE and SCSI CPU utilization for hard disks, even when
> DMA is enabled?

I can think of lots of reasons like average transfer size (smaller
transfers mean more setup by the CPU) or hardware scatter gather
support (multiple small transfers with little CPU interaction), but
none are fundamental differences between IDE and SCSI, rather
differences in the controllers and drivers you have.

-Jim