On 15 Dec 1999, Matthias Andree wrote:
> "Ingles, Raymond" <[EMAIL PROTECTED]> writes:
>
> > I'm happy with the NCR-based stuff, and I'd get a Symbios 53C875 in a
> > heartbeat, but from what I understand the sym53c8xx driver won't work
> > with a 53c825 (but would if I had a 53c825a). Is this correct? Can I use
> > both drivers at once? How bad would the performance hit be if I ran both
> > with the ncr53c8xx driver? Would that work?
>
> I never tried the sym + ncr 53c8xx drivers simultaneously, but I think
> you'd get along quite well with the ncr53c8xx for both of the chips. I
> never noticed the difference after switching my 53c875 from the ncr to
> the sym "enhanced" driver. Both are good quality. There was no actual
> reason for me to switch to the "enhanced" (sym) driver, I was just
> playing for fun, it worked well and so I stuck with it.
Some advantages of the sym53c8xx driver:
- Lower command latency (>10,000 Transactions/second shows undubitably
the differences, even with a 875)
- PCI 2.2 compliant. Never tells the chip to perform PCI self-mastering.
- Will (does with current version not yet user-available for the
sym53c8xx) support Ultra-3 DT transfer with the C1010 (devt done by
Pamela Delaney from LSILOGIC for the Linux sym53c8xx and independently
by me for the FreeBSD sym_hipd driver (sym53c8xx moved with some rewite
to FreeBSD).
- The maintainers (Pamela and me) prefer the sym* drivers. :-)
- The sym* drivers support the full 8K SRAM for the 896, 895A and future
Ultra-3 C1010, and handle phase mismatch from SCRIPTS.
- The sym* drivers support residuals (not yet validated under Linux but
code is there) and the Modify Data Pointer (not yet tested under both
sym*)
- There are plans for target mode implementation in the sym* drivers.
But, frankly, for common use, ncr* versus sym* should not make noticeable
difference for user.
Recommendation (from me :) ) is to use sym rather than ncr for chips
supported by sym (obviously if sym works for you).
There is no significant overhead in loading both drivers at the same time
other than the memory used for the code of the second driver (60Kb).
G�rard.
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
