About Triton DMA IDE vs. SCSI performance
On Thu, 8 Aug 1996, Douglas Bates wrote:
> For example, Bruce Perens mentioned on this list some time ago that
> the combination of the IDE controller in the Triton chipset with an
> EIDE drive is as fast as fast, wide SCSI. I found that the system I
> was using had the Triton chipset and enabled the Triton-specific
> optimizations in the kernel. The results were very pleasing.
I didn't say wide fast. On my system I am getting about 5 MBPS on IDE from
a Quantum Fireball drive, which is as fast as my narrow fast SCSI drive
on a narrow fast Adaptec AIC-7850 SCSI on the same motherboard (an IWILL
P54TS). The speed seems to be limited by the disk media, not by any
difference between IDE and SCSI.
From: Sherwood Botsford <sherwood@space.ualberta.ca>
> That would depend on how hard the disk subsystem is being hammered.
> SCSI's advantage is that it can handle multiple outstanding requests at
> the same time, while EIDE has to wait for one to complete before it can
> start a second one.
[...]
> (This allows the drive to start being clever and resort
> the queue'd I/O's for fastest return. E.g. If the head is on cylinder
> 7, and it gets a read request for cylinder 2000, and cylinder 500, it
> will do the 500 on the way to the one for 2000.)
Linux can do the ladder-seek algorithm in software for IDE drives, so
much of the intellegence of the SCSI drive is duplicated for IDE and
you get the closest blocks first just as you would with SCSI. The advantage
of the on-board disk cache is reduced, as Linux duplicates it with its own
cache on a faster bus.
No doubt SCSI works better for multiple devices on a single controller.
However, for my configuration this is probably moot - the Triton
provides two DMA capable IDE busses each with its own PCI bridge, and I
have one disk on each.
Bruce
Reply to: