On Sun, 17 Oct 1999 22:57:15 -0500 (CDT), you wrote:
>Alex,
> All the "old" CD-R faqs recommend NO other i/o activity during a burn -
>to the extent of disabling screen-savers, etc. I have a '97 copy of the
>Linux CD-R HowTo that recommends unplugging network connections! However,
>I understand that the lack of isolation of IDE activity is the main
>shortcoming of IDE writers. Maybe one of the hardware gurus will further
>enlighten us...
Doc & others,
I'm pretty sure there are two issues: First, the FAQ's were probably
written back when a 486 was state-of-the-art, so the processing power
was substantially lower than it is even with a low-end machine today.
Second, I think the SCSI adapter design vs. IDE is such that SCSI
drivers are much more efficient about writing data.
I recently wrote a batch of CD-R software on NT, and quickly got sick
of rebooting NT every other time I tried to write a CD (or fail
miserably.) Fortunately, it wasn't because of processor time, but
because the IDE/ATAPI drivers for NT suck ass (or there's something
else screwed up that I'm not aware of, which is probable.) I'm sure
it wasn't processing speed because I'm using a dual processor PII-450
with gobs of RAM. I got sick of the reboots (70% of my time) and went
to STS and spent $380 or so on a Teac CD-R55S, which came in a beat up
box and had obviously been returned several times (woo-hoo!) I
couldn't be happier now. (Now I just need a second one on this
machine.) It has never failed, even when I was kiling it with
lame-brained CD-R software bugs in my stuff.
I'm never going to use another IDE drive if I can avoid it.
More info, although it may not be relevant to Linux but may be
indicitive of other problems with IDE drives in general: I just had
four machines set up at Photodex with IDE CD-Rs, and some of them have
significant problems reading CDs for things like sofware installation.
They're all Windows machines, though, so it could just be the Windows
drivers. I suspect the IDE drives will be more sensitive to system
load.
I'm very intersted to hear form a hardware guy out there who could
enlighten us about the design differences between SCSI and IDE.
Perhaps IDE drivers are required to spend a certain amount of time
polling, which would explain the processor utilization and problems
with tasking. It would also explain why SCSI is so much more
effecient. I'm pretty sure SCSI almost always is DMA driven with
queues.
(Boy...can you tell what Photodex is about to do in CompuPic?) <grin>
Best Regards,
Paul Schmidt ([EMAIL PROTECTED])
Photodex Corp.* 1106 Clayton Lane 200W * Austin, TX 78723
(512)406-3020 Voice * (512)452-6825 Fax * www.photodex.com
---------------------------------------------------------------------------
Send administrative requests to [EMAIL PROTECTED]
