At 13:43 01-03-98 +0900, Lance Cummings <[log in to unmask]> wrote:
>Perhaps you can tell me if the ability of ultra to double the clock
>rate means that ultra would probably support higher bus speeds...
>such as 75 mhz or 83 mhz.

When you overclock the system bus from 66 mhz to 75 mhz or 83 mhz, as
you know, this changes the PCI bus from 33 mhz to 37.5 mhz or 41.5 mhz.
Some PCI cards cannot handle this.

From reading mailing lists and Usenet messages, I hear that the Adaptec
2940U and 2940UW seem to be more overclockable than the non-ultra 2940
and 2940W. This might be because the ultra cards are of a more recent
design or, as you surmise, maybe the ultra capable cards are more
tolerant since they are designed to transmit data at a faster frequency:
or maybe because the BIOS on the ultra models in the 2940 series is more
recent. (I've seen people claim that a BIOS upgrade let them overclock
a card that was previously not overclockable.) There are claims that
replacing the clock crystal on the 2940xx will allow it to be
overclocked. (I have an email that I saved concerning this.) All this
may just be folklore. Certainly the evidence is anecdotal, and there
are no guarantees.

Maybe the Adaptec cards are engineered to closer standards than some
other brand SCSI cards, thereby making them less overclocking friendly.
(But Adaptec does have a very good reputation for stability when run
within specs.)

SCSI controllers which use Symbios chips are supposed to be easier
to overclock than Adaptec's PCI controllers.
SymBios (formerly NCR) 53C810, 53C825, 53c860 and 53c875 PCI SCSI
controllers, Diamond Fireport 40 (53c875), ASUS SC-200 (53c800)
no BIOS, ASUS SC-875 (53875), Tekram 390U (53c875), Data Technology
and Tyan cards. Note that Adaptec is in the process of acquiring
Symbios.

If I were going to buy a new SCSI controller, I would go for an
Ultra2 controller like the Adaptec 2940U2W described at
<http://www.adaptec.com/products/overview/aha2940u2w.html>
Benefits include longer allowable cable lengths, double the bandwidth,
backwards compatible, driver support for Win95 and NT4, Unix soon...

Finally, I think that any speed increase from overclocking would be
useless if it came with corrupted data or system instability. Better
to wait 6 months (or whatever) for the upcomming 100 mhz bus
motherboards where the PCI bus should be run asynchronously from the
system bus, i.e. at the normal 33 mhz. That would make this whole
speculation mute.

Regards,
Bill