On 12 Oct 2000, Thierry Vignaud wrote:
> >
> > What's more, IDE suffers from not having TCQ (well, AFAIK, ATA 100 will have)
>
> yes but on a UP box, the scsi command overhead makes ide looks better.
>
What do you mean by command overhead? Disk access time?
I did an experiment at home, where I have two 9GB IBM SCSI disks, one CD writer
and a DVD at home on a Tekram DC390U2W (sym53c815 chipset). On my UDMA66 IDE
(which is recognized under Linux - VIA KX133 rocks), I had only a Maxtor 20GB.
hdparm gave clear advantage to raw disk reads to IDE: 26 MB/sec against 20
MB/sec for SCSI (but cached writes were faster on SCSI!). But when I did a real
world test (ie, tar cf /dev/null /usr/src/linux), SCSI won hands on (only 60%
of the time than what IDE asked), added to the fact that system CPU time taken
was close to 0 (more than 7% for IDE).
Don't rely on raw performances ;)
--
Francis Galiegue, [EMAIL PROTECTED]
"Programming is a race between programmers, who try and make more and more
idiot-proof software, and universe, which produces more and more remarkable
idiots. Until now, universe leads the race" -- R. Cook
