Date: Tue, 2 Feb 1999 10:18:49 +0000 (GMT) From: Doug Rabson <dfr@nlsystems.com> To: Thor Lancelot Simon <tls@rek.tjls.com> Cc: netbsd-alpha@netbsd.org, freebsd-alpha@FreeBSD.ORG Subject: Re: (PROBLEM SOLVED) Re: restore to srm that doesn't.... (fwd) Message-ID: <Pine.BSF.4.01.9902021014310.10768-100000@herring.nlsystems.com> In-Reply-To: <19990202050109.A19540@rek.tjls.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 2 Feb 1999, Thor Lancelot Simon wrote: > On Tue, Feb 02, 1999 at 09:39:48AM +0000, Doug Rabson wrote: > > On Mon, 1 Feb 1999, Matthew Jacob wrote: > > > > > > > > An interesting FYI about CIA. Jay works at Digital.... > > > > > > > > > > > > ---------- Forwarded message ---------- > > > Date: Mon, 1 Feb 1999 19:22:39 -0800 > > > From: Richard Henderson <rth@cygnus.com> > > > To: mjacob@feral.com > > > Cc: linux-alpha@vger.rutgers.edu > > > Subject: Re: (PROBLEM SOLVED) Re: restore to srm that doesn't.... > > > > > > On Mon, Feb 01, 1999 at 05:15:40PM -0800, Matthew Jacob wrote: > > > > cia0: using BWX for PCI config access > > > > > > Beware. Jay Estabrook told me the CIA-2 on the PC164 is buggy > > > with BWX I/O. The first core chipset to get it right was PYXIS. > > > > > > I don't know in what form the bugs appear -- I just took his word for it. > > > > > > > isp1 at pci0 dev 6 function 0 > > > > isp1: Board Revision 2100, loaded F/W Revision 1.15 > > > > isp1: Adapter WWN 0x200000e08b003c1f > > > > isp1: Firmware State Config Wait -> Waiting for AL/PA > > > > isp1: Firmware State Waiting for AL/PA -> Wait Login > > > > isp1: Firmware State Wait Login -> Ready > > > > isp1: Loop ID 113, ALPA 0x23 > > > > > > What's this card? > > > > > > > sio0 at pci0 dev 8 function 0: Intel 82378ZB System I/O (SIO) (rev. 0x43) > > > > > > Yep, definitely a PC164. Wonder why the SRM isn't reporting the > > > right thing? > > > > > > Would you annotate get_cpuinfo() to report the numeric sys_variation? > > > Perhaps my ident data is off by one. > > > > We only use BWX for machines with PYXIS. As far as I know, all EB164 type > > machines with a subtype >= SV_ST_ALPHAPC164LX_400 should have PYXIS (this > > is the way NetBSD detected them and we do the same). > > Early Pyxis chipsets have fairly dreadful DMA bugs, too, IIRC. I distinctly > recall a few people who could basically never get acceptable performance from > any operating system on their pc164lx motherboards, even Digital UNIX. Both BSD's try to detect this and disable prefetching on the PCI bus to work around it. One of my machines (a 433au) has this bug and is stable with the workaround. I don't know what kind of performance hit is involved since the machine is fast enough for my purposes :-) > > Interesting tidbit: a PC164 (no LX) with all the memory slots full has more > memory bandwidth than a PC164LX, because the PC164LX can't interleave its > SDRAM. It's also a lot less likely to have various dreadful bugs... > nb01.netbsd.org is a pc164 which chugged along as a fileserver for me for > months with great performance and no notable problems. I kinda wish they'd > kept making the plain PC164 a little longer... Cool! And I thought the memory bandwidth on my PC164LX was good... -- Doug Rabson Mail: dfr@nlsystems.com Nonlinear Systems Ltd. Phone: +44 181 442 9037 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-alpha" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.01.9902021014310.10768-100000>