asami at dt.wdc.com
Tue May 23 10:31:22 PDT 1995
Steve Finch of Silicon Systems wrote:
> When looking at Fast-40, I think it is time to also look at the overhead!
Steve, thank you for bringing up the subject. Your throughness always
impress me. And I do agree with what you said. That has been my sore point
on SCSI for the longest time, too.
But just to be a devil's advocate, let me bring up the counter-argument, and
I hope someone can come forward and break it.
The timing parameters such as arbitration time, bus settle time, etc. which
seem very large (high hundreds of nanoseconds to microseconds) compared to
data cycles of Fastxx SCSI (currently 50nsec data cycle, going on 25nsec),
but they are defined in terms of 30 meter cable and physics.
I have the original John Lohmeyer spreadsheet defining these timings, and
they have to be of that order of magnitude if we need the cable length.
One of the driving point for the SPI-2 (pointed out as one of the weakness
against FC) is the cable length. To compete against newer interfaces (why?
resistance is futile?) 6m limit single ended, 3m for Fast-20 (with severe
load limitation at that) is not desirable. One of the motivation to go
differential is to extend the cable length limit back to 25m, or possibly
more. You can do that with Fast-20 today, if you are differential, but the
driver requirement is such that it is too power hungry to be integrated on
a monllithic silicon.
And another one is (you might have guessed it), the backward compatibility.
Up to now, Fast-20 devices and Fast and Slow SCSI devices can live happily
on the same bus (theoretically!). If we cut down on these timings, we tend
to cut off that inter-operability. Well, this may not be such a strong
argument since the old devices don't talk LPD.
So! We need the balance. What is our optimum goals for overhead and
cable length? Backward compatibility?
_|_|_| asami at dt.wdc.com
_|/ _| WESTERN DIGITAL (714) 932-7621 : Voice
_|_|_| (714) 932-6496 : Fax
More information about the T10