home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.sys.sgi
- Path: sparky!uunet!europa.asd.contel.com!darwin.sura.net!mips!odin!sgihub!zola!zuni!anchor!olson
- From: olson@anchor.esd.sgi.com (Dave Olson)
- Subject: Re: Memory upgrades for R4000 Indigos
- Message-ID: <npq4its@zuni.esd.sgi.com>
- Sender: news@zuni.esd.sgi.com (Net News)
- Organization: Silicon Graphics, Inc. Mountain View, CA
- References: <31535@adm.brl.mil> <noj5dkc@zola.esd.sgi.com> <np0l1qk@rhyolite.wpd.sgi.com>
- Date: Sun, 26 Jul 92 20:27:17 GMT
- Lines: 58
-
- In <np0l1qk@rhyolite.wpd.sgi.com> vjs@rhyolite.wpd.sgi.com (Vernon Schryver) writes:
-
- | In article <noj5dkc@zola.esd.sgi.com>, portuesi@tweezers.esd.sgi.com (Michael Portuesi) writes:
- | > ...
- | > An R4000-based Personal IRIS, if it existed, would
- | > probably cost considerably more than the Indigo R4000
- | > due to these manufacturing constraints, and hence
- | > would have few, if any, buyers.
- |
- | Please be careful. I don't think this is accurate. A bunch of
- | customers out there are probably fuming, thinking we think they are
- | really stupid.
-
- He's right. It was *incredibly* painful getting the 35 to a point where
- it worked reliably. The original design just wasn't oriented towards the
- amount of air flow the 35 needed, and in the places it needed it.
-
- | Yes, I don't doubt it would not work to just jam an R4000 into the
- | 4D/20 case, as was done with the 20MHz R3000 and 38 (35?) MHz R3000.
-
- We completely redesigned the electronics module and added scatter fans
- or small auxilliary fans (depending on when the system was made) in order
- to get adequate air flow for the 35. That is hardly cramming it into
- the same box, although that *was* the original idea.
-
- | I know a just a little of the heat and other problems with the 4D/35 from
- | a certain VME board. The airflow in the box was fine for the 12MHz
- | 4D/20, but is a long way from good enough for the R4000. So what?
-
- It wasn't even good enough for a 36 Mhz r3000A; we would have shipped
- it at 40 Mhz if we could have made it reliable. A very few carefully
- hand adjusted systems run reliably at 37 or 38, but none at 40, unless
- they were open on a bench with a fan on them. Almost all of the problems
- were airflow, not chip margins.
-
- | If you step back two paces, you'll see there is little difference
- | between the Indigo box and the PI box. One is a little wider. The
- | other is a little deep and much taller. Oh, the colors are different.
- | So what?
-
- The design for the airflow, the (much!) lower wattage powersupply, and
- the physical design of the Indigo helped a lot, believe it or not. Yes,
- if we redid the skins, the powersupply, the whole chassis, and the
- electronics module, we could put an r4k into a PI (but then it wouldn't
- be a PI anymore!), but that would be far more money (and people's
- time!) than anybody could justify, when the Indigo could be made to
- work much more easily. There just wasn't any point. For the
- relatively small number of customers who want VME on r4k, there is
- crimson, at not much more than an r4k PI would cost after we put all
- that work into it.
-
- Besides that, the VME interface is no longer part of the chips that we
- use in the r4k Indigo, so we would have to go (re)design another chip on
- top of everthing else...
- --
- Let no one tell me that silence gives consent, | Dave Olson
- because whoever is silent dissents. | Silicon Graphics, Inc.
- Maria Isabel Barreno | olson@sgi.com
-