[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: growisofs will not write at 8x with NEC 3540A



Hi,

> I have an NEC 3540A along with Ritek Ridata 8x DVD-R's.
> If i burn a DVD iso in windows with nero I get true 8x speed.
> When I burn in linux with growisofs, my speeds vary from 1.5x-4x
> which is horrible.  Is there no 8x support yet?

As far as i know growisofs leaves the speed setting as
proposed by the writer (unless you use the -speed= option).
What speed is reported in its own start message ? Like in:
  About to execute 'builtin_dd if=/dev/fd/0 of=/dev/sr0 obs=32k seek=0'
  /dev/sr0: "Current Write Speed" is 4.1x1385KBps.


Varying speeds indicate rather a problem with the feeding
of growisofs than with the writer. One should try to get
the throughput steady before trying to improve speed from 
4x to 8x.

Throughput obstacles known to me:

- Writing on the fly with a system where mkisofs cannot deliver
  a steady output stream of 4x speed.
  A fifo buffer in a separate process or thread can work wonders
  if the *average* output speed of mkisofs is high enough for
  full speed.
  The input of mkisofs can be hampered by the activities of
  growisofs and the burner (what do i expect from a mainboard
  which costs less than 50$ ?). In that case a fifo might help
  or help not.

- Disadvantageous IDE address combinations and DMA settings can
  slow down the transfer from disk to writer independently of
  intermediate processes. 
  I recently had the case where disk=hda + writer=hdb + highspeed DMA
  led to i/o errors which looked exactly like media failures but
  were clearly correlated to a special gesture of my own software.
  (To my luck that gesture was absolutely incapable of influencing
   burn success directly. So the storage location of some intermediate
   data was quickly spotted as the culprit.)


In your situation i would also try cdrecord-ProDVD wether it is willing
to set higher speed resp. wether it is telling some information about
the problem. cdrecord-ProDVD got a builtin fifo. So you can check
wether this would help with the varying throughput. 

I would also try both programs wether they can burn with steady 
throughput if fed by cheap dummy data. Like :

   dd if=/dev/zero bs=1M count=4482 | \
   growisofs ... -Z /dev/sr0=/dev/fd/0
resp.
   cdrecord-ProDVD ... fs=32m speed=8 -   
   (option -force might be worth a try, too)


Have a nice day :)

Thomas



Reply to: