Paul
Mon Apr 21 12:15:15 PDT 2008
mjs wrote:
> "Robert Moir" <usenet@REMOVE2EMAILrobertmoir.com> wrote in message
> news:O1Zz$A8oIHA.4292@TK2MSFTNGP04.phx.gbl...
>> Let's just clear something up. Are you saying you're having trouble
>> copying _data_, in other word dragging and dropping files, or are you
>> importing DVD film into something? Because the time would be about right
>> for the latter. For the former, I'd agree with Pegasus.
>
> I'm grabbing a bunch of data files from a DVD I burned myself for backup
> purposes, and copying them back onto my HD. They're mostly wav files. I'm
> just copying them.
>
There are a couple things to check, and it made for an interesting
experiment here (fixed mine :-) ).
Optical drives now, could be connected via IDE or SATA. I happen to
have a CDROM on this machine, and it is IDE (ribbon cable).
I checked Device Manager, and while my primary interface was
"DMA if available", my secondary interface (the ribbon cable with
just the CDROM on it), was in "PIO" mode.
I fired up a copy of Sisoftware Sandra Lite, and used the CD/DVD benchmark.
It does a linear read of the data on the disc. It seems to want real
data on the disc, so I used my copy of Knoppix (700MB stored on CD) as a
test disc. When I ran the Sisoftware benchmark, it was sloped at the
beginning of the CD, but flattened out at about 3.6MB/sec or so for
the second half of the disc.
That is consistent with PIO mode transfer (polled mode). So in fact,
the interface mode selected for the CD drive, was holding it back.
I set the Device Manager properties to "DMA if available" for the
secondary interface, and rebooted the computer. (This is Win2K I'm using,
so YMMV.)
After reboot, I reran the benchmark. This was more like it. The graph
had a steadily rising transfer rate, across the disc. Just like the other
graphs included as sample results, included with Sandra. Since the
graph was tilted, and didn't have any flat sections, I could conclude
from that, my test was "media" limited and not "cable" limited.
2.45MB/sec 16.4X beginning of 700MB CD
5.85MB/sec 39.0X end of 700MB CD
seek 106msec full stroke estimated
(1X = 150KB/sec for CD, a different conversion factor is used for DVD)
Transfer performance on a device is affected by two things. Transfer
bandwidth (which may be a function of which track the device is
accessing). And seek time. Seek time is an issue, if you're randomly
accessing files.
For example, say a CD had 100000 small 2KB files on it. And you randomly
transferred them from the CD, to your hard drive. The transfer would be
dominated by seek time. In my case, I'd have to wait up to 106 msec for
the head of the optical drive, to get to the new track. The average
wait might be 34 msec. Since the files are small, in this example case,
it wouldn't matter whether DMA or PIO transfer mode was used, because
the transfer interval, compared to the seek time, is so small.
The best way to transfer data from a storage device, is sequentially. For
example, if I zipped the contents of my hard drive, and made one big zip
file on the CD/DVD, then during the copy operation, the transfer would be
a largely linear transfer, with not a lot of seek component to it. Thus I
could get a rate between 2.45MB/sec and 5.85MB/sec. But if fetching 2KB
files all over the disc, I'd be lucky to see 20KB/sec as a transfer rate,
due to the seek operations for each 2KB file.
By looking at the linear transfer curve, you can tell whether transfer
is media or cable limited. And whether things are working decent or not.
You can get the Lite (free) version here. This version is one version later
than my copy.
http://www.sisoftware.co.uk/index.html?dir=dload&location=sware_dl_3264&langx=en&a=
HTH,
Paul