Date: Sun, 13 Aug 1995 14:41:46 -0700 (PDT) From: "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> To: pete@puffin.pelican.com (Pete Carah) Cc: hardware@freebsd.org Subject: Re: Upgrade to my machine Message-ID: <199508132141.OAA11034@gndrsh.aac.dev.com> In-Reply-To: <199508131613.JAA01053@puffin.pelican.com> from "Pete Carah" at Aug 13, 95 09:13:12 am
next in thread | previous in thread | raw e-mail | index | archive | help
> > In article <199508130850.BAA09255@gndrsh.aac.dev.com> Rod writes: > ..... > >> I have a Hawk and a Barracuda and recommend them both with the following > >> proviso: The Barracuda runs very hot. I have it mounted in a server case > >> containing 3 fans (5 if you count the power supply and CPU.) One of the > >> fans blows directly across it. Still warm to the touch but not worth getting > >> paranoid over. Performance wise it rocks. > > >And that ``proviso'' is one many are not willing to live with. If a drive > >requires that type of external cooling air blown at it, then something is > >seriously wrong. > > Well, the official word from SGI is to not mount Barracuda drives internal > to any SGI except the large Challenge/Onyx rack for this reason. > (old indigo, indy, all older machines; actually they may have improved > cooling on indigo2 but I don't remember...) There are a couple of other > drive lines in that list but I don't remember which either. And the ``official word'' from AAC is not to run a Barracuda at all. :-) > >Read my .signature, and then think about what I said, and you may fully > >understand why I said it :-) :-) > > Barracuda drives seem just fine in the "appropriate" external case... Try that as a sales pitch to a customer about to buy a $5k Pentium system who needs 4G of high speed disk. It doesn't fly very well! > We have seen the usual > problem with non-ideal scsi cables; the SGI indy's are especially susceptible > to this (along with the older 1542C's that Rod has already commented on, > and whatever else). (actually all SGI's are susceptible to this; there > seems to be something about the WD scsi chipset that they use...) It has very little to do with the SCSI controller chip itself, it is mostly due to signal edge rates which are directly attributable to the type of buffer chip you use between the controller chip and the bus itself. Adaptec, like many other vendors have gone to the Dallas 2107AS active terminator/scsi bus driver which happens to have a very fast edge rate. Well, very fast edge rates and low quality cables don't mix very well. The real failure here is caused by the incomplete specification called ``SCSI-II''. It only specifies a ``maximal'' signal rise/fall time, when doing real bus and transmission line design this is only 1/2 of the data needed to insure you design a reliable bus. Minimal signal rise/fall time is a very important parameter when dealing with controlled impendence bus design, and one that is all too often overlooked. :-(. The Dallas chips work _great_ if you happen to be driving the proper 110 ohm actively terminated, properly shielded and seperated transmission lines that the SCSI-II specification recommends. However, they cause serious crosstalk and ringback on a typical SCSI-I class 130 ohm, single shielded cable with passive termination :-(. Hook a scope up to the later and just take a quick peek, you'll see the problem! -- Rod Grimes rgrimes@gndrsh.aac.dev.com Accurate Automation Company Reliable computers for FreeBSD
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199508132141.OAA11034>