Re: sym53c875E-0: PCI clock seems too high

Andreas Bombe (andreas.bombe@munich.netsurf.de)
Tue, 28 Sep 1999 09:47:34 +0200


On Sun, Sep 26, 1999 at 04:19:43PM +0100, Paul Flinders wrote:
>
> Andreas Bombe <andreas.bombe@munich.netsurf.de> writes:
>
> > On Sat, Sep 25, 1999 at 11:08:26PM +0200, Gerard Roudier wrote:
> > > Overcloking a CPU that is in fact sold with an underclocked specification
> > > is one thing. Overclocking a BUS and then everything that connect to it by
> > > more than 10% was not imaginable by me. By the way, I think that people
> > > that overclock PCI 33 Mhz buses up to 41MHz or more are just foolish and
> > > plain idiot, but I don't care of them.
> >
> > Even though bus overclocking may work for many devices, SCSI
> > controllers are known to react very allergic to that. Overclocking the
> > bus while using SCSI is therefore exceptionally stupid, given that data
> > integrity could suffer.
>
> First off let's remember how CPU's are manufactured. Typically there's
> no difference between, say, a 433Mhz chip and a 333Mhz chip - the
> likelyhood is that they came off the same production line and the only
> difference is that the faster chip passed a (quite short) set of tests
> at that speed.

Yes, CPUs are manufactured for a wide range of clocks. And CPUs are
more or less self contained, the internal operation is decoupled from
the FSB (except for the multiplier of course). Therefore they are
quite easy to overclock.

> The same applies to all the components in the system - all are
> designed to work with an "acceptable" error rate at a given speed but
> some will work at higher speeed, however, unlike CPUs there's not
> usually a higher value attached to faster parts.

But PCI chips are manufactured solely for a 33MHz bus clock, there is
no excess capability planned. Overclocking may work, but with different
success. Graphics cards may give pixel errors when run too fast, or
they don't. But they won't kill your data.

However, according to c't (which did overclocking tests on PCI), SCSI
controllers can almost be expected to fail. Probably because they
drive some internal clock from PCI or the chips are clocked from PCI
and, unlike CPUs, are built only for 33MHz. SCSI controllers are
apparently coupled to the bus clock tighter than other devices.

Other types of cards probably make less problems.

-- 
    Andreas E. Bombe <andreas.bombe@munich.netsurf.de>
    http://home.pages.de/~andreas.bombe/
RSA 0x886663c9  30 EC 09 73 84 7B 55 83  C4 7A 91 D9 9D C5 4B B0
DSA 0x04880A44 72E5 7031 4414 2EB6 F6B4  4CBD 1181 7032 0488 0A44

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/