On Wed, Aug 13, 2003 at 06:27:57AM -0400, Sean Davis wrote:

> All I have connected right now is:
> sd0 at scsibus1 target 0 lun 0: <WDIGTL, WDE4550 ULTRA2, 1.22> disk fixed
> sd0: 4340 MB, 6932 cyl, 6 head, 213 sec, 512 bytes/sect x 8890000 sectors
> sd0: sync (100.00ns offset 12), 16-bit (20.000MB/s) transfers, tagged
> queueing
> cd0 at scsibus1 target 3 lun 0: <PLEXTOR, CD-ROM PX-40TW, 1.04> cdrom
> removable
> cd0: sync (100.00ns offset 12), 16-bit (20.000MB/s) transfers
> 
> Both on a QLA1020. FastWide doesn't seem to quite cut it. 
Lets do some math. We asume that the geometry reported by the drive 
fits approximately the real geometry. (I am sure the drive does ZBR.)
Further we asume 7200 RPM. That is much for an old 4 GB disk. So we get 
213 sec/track / 2 sec/kB * 7200 RPM / 60 s/min = 12,780 kB/s
maximum sustained data rate for one track. If there are head movements 
needed this wil be lower. Even with 10 kRPM we get 17,750 kB/s 
what is less then the 20 MB/s of the controller. 

> I used to use
> Ultra wide with an equal-size hard drive on a pentium 166 (the machine in
> question is an alpha 21164 266mhz) and it was quite a lot faster than this.
Maybe this drive has more RPM (lower latency), more cache, write back cache, 
or ... and your application load may differ. Only on a drive with a big cache
a faster interface will make a difference. 

IMHO. 
-- 



tschüß,
         Jochen

Homepage: http://www.unixag-kl.fh-kl.de/~jkunz/