From owner-freebsd-hackers Sat Nov 30 15:30:42 1996 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.7.5/8.7.3) id PAA14893 for hackers-outgoing; Sat, 30 Nov 1996 15:30:42 -0800 (PST) Received: from covina.lightside.com (covina.lightside.com [207.67.176.1]) by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id PAA14888 for ; Sat, 30 Nov 1996 15:30:40 -0800 (PST) Received: from bebox (hamby2.lightside.net [207.67.176.18]) by covina.lightside.com (8.8.3/8.8.3) with SMTP id PAA04499 for ; Sat, 30 Nov 1996 15:30:31 -0800 (PST) Message-Id: <199611302330.PAA04499@covina.lightside.com> To: hackers@FreeBSD.org Subject: FreeBSD/Alpha (was Re: COMDEX trip report) Date: Sat, 30 Nov 96 15:29:21 From: "Jake Hamby" Reply-To: jehamby@lightside.com X-Mailer: BeMail [version 1.1] Sender: owner-hackers@FreeBSD.org X-Loop: FreeBSD.org Precedence: bulk Sorry if this is a duplicate message. I'm using BeMail, a prerelease mailer on a prerelease OS, and it appears that I just a blank message to hackers so I'm resending it... >> platforms@freebsd.org exists for just this purpose. 8) > >True, but perhaps more focused mailing lists are in order. >I'm not sure, for example, that I'd want to know every detail >of the FreeBSD/MIPS project, but I would be quite interested in >the FreeBSD/ALPHA project. That reminds me: Scott McNealy at Sun said a few years back that only three CPU's would be commercially significant in the future: SPARC, x86, and PowerPC. Then, they ported Solaris 2.5.1 to all three. However, the latest rumor is that Sun will stop supporting PowerPC (I don't know *anyone* who uses Solaris/ppc) and go back to SPARC and x86. The moral of that story, as I see it: 1) Don't try to support more platforms than you can handle. I think that more than two different architectures would be more than FreeBSD could support with the number of active developers right now. So the question is whether that second architecture should be Alpha. 2) Choose platforms with large volume. If Scott McNealy's comment accurately reflects market share (and I believe it does), a logical choice would be SPARC or PowerPC. 3) Choose a platform which you have a chance of infiltrating. In other words, people with PowerPC's were most likely already using AIX, therefore Solaris didn't have enough market share to be profitable for a commercial vendor. Well these three statements contradict each other! So, which is the logical next platform to support? I'd prefer PowerPC, but Alpha is also logical. Here are the pros and cons: Alpha pros: 1) "Coolness" reputation as the fastest mass-market CPU. 2) Already supported by Linux, so a FreeBSD port helps our "rivalry". 3) Many PCI drivers would carry over from the x86 port with some changes. Alpha cons: 1) Lousy price/performance ratio. 2) Difficulty of porting to 64-bit (or making a 32-bit version and face ridicule from Linux/Alpha users!) PowerPC pros: 1) Excellent price/performance (up to twice that of Intel). 2) Largest volume RISC chip. 3) PCI drivers would carry over from the x86 port with some changes. 4) Also supported by Linux, but not advertised much. 5) SMP systems readily available, a PowerPC performance advantage over Alpha (dual PowerPC's are cheaper, and also faster, than a single Alpha). 6) If you really want the world's fastest CPU, the 533MHz Exponential PPC clone will be shipping in volume next year. PowerPC cons: 1) Which systems to support? PowerMac-based? OpenFirmware-based? BeBox? 2) SMP support would be difficult, but a great selling point vs. Linux. Okay, that's how I see the issue. I'm admittedly biased, as a BeBox owner, but I do believe the price/performance advantage and high volume of PowerPC is a clear benefit over supporting the sexy, but overpriced Alpha architecture. Just my $0.02... -- Jake