IDE vs SCSI drives

[Kevin Buettner]

On Oct 18,  7:53am, Rusty Carruth wrote:

> > SCSI subsystem driver Revision: 1.00
> > (scsi0) <Adaptec AIC-7890/1 Ultra2 SCSI host adapter> found at PCI 0/14/0
> > ...
> 
> unfortunatly you left out the total hosts line ;-)

Do you mean a line that says something like:

    scsi : detected N SCSI disks total.

??

If so, I'm seeing a message like this on a box with a 2.2.15 kernel,
but I'm not seeing it on the machine in question which is running
2.2.4-test9.

> > Comments?  In particular, I'd like those SCSI advocates to speak up
> > and let me know what I'm doing wrong with my SCSI drive.  (I'd hate to
> > think that I've been paying more money all of these years for less
> > performance.)
> > 
> do you have anything else on the scsi chain?

I have a scanner on the SCSI chain too.

> The slowest device will (I understand) drag all the other ones down
> with it.  SO if you (like me) have an old scsi1 cdrom drive that
> only talks the slowest protocol you'll get stuck down at what,
> 10MB/sec?  Those who know for sure feel free to correct me if I'm
> 'worng' (sic)

Interesting.  I would hope that's not the case, but I'm not a SCSI
expert.  Perhaps someone else could address this?

In any event, here's the latter part of /proc/scsi/aic7xxx/0:

(scsi0:0:2:0)
  Device using Narrow/Async transfers.
  Transinfo settings: current(0/0/0/0), goal(255/0/0/0), user(10/127/1/0)
  Total transfers 0 (0 reads and 0 writes)


(scsi0:0:6:0)
  Device using Wide/Sync transfers at 80.0 MByte/sec, offset 31
  Transinfo settings: current(10/31/1/0), goal(10/127/1/0), user(10/127/1/0)
  Total transfers 702533 (427657 reads and 274876 writes)

Note that the listed transfer methods are different for the scanner
vs. the disk.  (Yes, the disk is the one that says that its
transfers are at 80.0 MByte/sec.)

> (And I'm surprised, because the same kind of tests that I've run on
> my scsi systems always show scsi better than IDE...

In this case, I think it may be due to the fact that I'm using a one
year old SCSI disk.  The product literature on Quantum's website
indicates that the average sustained transfer rate is 12 MB/sec and
that is roughly what hdparm -t reported.  (Actually, the hdparm
results were slightly better.)

I'm still surprised at the performance of the IDE drives.  These are
large drives (45GB), however, and I read somewhere that the increased
data density of these large disks also results in higher sustained
transfer rates.

> Especially since you've shown that there are apparently no bandwidth
> bottlenecks after the ide or scsi...  hey, wait.  What kind of card
> are these?  8-bit, 16-bit, PCI, ???  If the hardware interface
> between the controller and the memory is different between the two
> types of controller you can see important effects there.)

The SCSI controller is on the motherboard.  The IDE controller is a
card plugged into one of the PCI slots.  Both of these show up in
/proc/pci, so they're definitely PCI devices:

PCI devices found:
  ...
  Bus  0, device  14, function  0:
    SCSI storage controller: Adaptec AHA-2940U2/W / 7890 (rev 0).
      IRQ 10.
      Master Capable.  Latency=64.  Min Gnt=39.Max Lat=25.
      I/O at 0xe800 [0xe8ff].
      Non-prefetchable 64 bit memory at 0xfebff000 [0xfebfffff].
  Bus  0, device  16, function  0:
    Unknown mass storage controller: Promise Technology, Inc. 20262 (rev 1).
      IRQ 11.
      Master Capable.  Latency=64.  
      I/O at 0xeff0 [0xeff7].
      I/O at 0xefe4 [0xefe7].
      I/O at 0xefa8 [0xefaf].
      I/O at 0xefe0 [0xefe3].
      I/O at 0xef00 [0xef3f].
      Non-prefetchable 32 bit memory at 0xfeba0000 [0xfebbffff].

Thanks for your feedback.

Kevin