home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sgi
- Path: sparky!uunet!gatech!emory!sol.ctr.columbia.edu!venezia!penev
- From: penev@venezia (Penio Penev)
- Subject: Re: SCSI floppy disk on INDIGO: Why is it so slooow?
- References: <unepq0o@zuni.esd.sgi.com>
- Sender: nobody@ctr.columbia.edu
- Organization: Rockefeller University
- Date: Mon, 11 Jan 1993 09:21:39 GMT
- X-Newsreader: TIN [version 1.1 PL6]
- Message-ID: <1993Jan11.092139.28106@sol.ctr.columbia.edu>
- Reply-To: penev@venezia.rockefeller.edu
- X-Posted-From: venezia.rockefeller.edu
- NNTP-Posting-Host: sol.ctr.columbia.edu
- Lines: 24
-
- Dave Olson (olson@anchor.esd.sgi.com) wrote:
- :
- : No! There *are* s/w buffers on the host. That doesn't help. You
- : have an embedded drive controller that reads/writes a sector from the
- : drive to/from the controller ram, then the controller turns around
- : and gives the data to the host (for a write), or is ready to accept
- : another command (for a read). All of that takes time, and by the time
- : the next sector is passed to the drive, the drive has already passed
- : the next sector, so you lose a rev. This is the case whether you
- : ask for one sector at a time or 1000. Basicly the embedded controller
- : was designed to be very cheap (as were the 2-3 others available at
- : the time, including NCR's), which translates to slow.
- :
-
- As far as I remember, the controller has a mode of reading a series of tracks with
- one operation. On the PC the transfer of the data is done by DMA. You specify to
- the DMA controller the address and the length and to the FDC the startin sector
- and the number of sectors. I do not know what's the connection of the controller
- to the SGI boxes (SCSI?), but suppose that analogous interface would work.
- I remember 300Mbits/sec on the 720Kb floppies and 500 on the 1.2M ones.
- This is of course the upper limit.
-
- -- Penio.
-
-