> Volker, are you using ide-scsi or ide-cd? I always use ide-scsi with burners and so far never had trouble with it. It's also required for Andy's random write patch.
Btw you can change that at run-time by reloading modules ide-cd and ide-scsi (need to rmmod some other modules first). Both grab all available devices when loaded, unless you use ignore=hdb,hdd etc on the first one. > I'm actually beginning to wonder if the issue is partially my ide cable > itself. Oh wow, they cost 4EUR here (proper high-density ones), at the first hint of trouble the scissors go through the old one ;))) They are actually more fragile than one generally assumes, especially as one is forced to pull them off at the cable instead of at the connector. > anything to do with the CPU itself. It very much sounds like a scheduling > issue, really, and I'd like to find some other people with the same issue to Can't comment, but Linux does seem to have too much trouble with IDE hardware. > >to agree 100% with Andy's assessment of the situation. He gave sound > >reason for his conclusion. I am still very interested in your detailed > >technical explanation on why a fast CPU with 0% load only manages a 0.4x > >DVD burn speed because of a lack of enabled DMA. > > Let me comment this by telling you that I definitely did see exactly this > behavior on a system that cannot properly do DMA. > The transferrate (after BurnProof had been enabled) was noticable below 1x, > the drive was a Pioneer A06. If you write something which _may_ be right, and of which everyone, including yourself, expects that it is a very unusual set of circumstances, it would have helped a lot if you had mentioned that you have *actually observed* such a [EMAIL PROTECTED]@# system. > >good sign). The read speed for the disk ends up topping out at around 3MB/s > >with the CPU using about 80%-90% CPU time in kernel space. A test of reading > >a DVD-R 4x shows around 4.5MB/s with the same CPU usage though a lot more > >stutter in apps. > > This looks as if there is no DMA :-( Agreed (unless the test was with readcd -c2scan). A good way to find out about the effectiveness of DMA is to observe the cpu load while reading, cat /dev/cdrecorder >/dev/null will do. Toggling the dma bit with hdparm should cause noticable changes in cpu load. Volker -- Volker Kuhlmann is possibly list0570 with the domain in header http://volker.dnsalias.net/ Please do not CC list postings to me.
