Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 01 Dec 96 11:24:05 
From:      "Jake Hamby" <jehamby@lightside.com>
To:        Mark Mayo <mark@quickweb.com>
Cc:        "Jordan K. Hubbard" <jkh@time.cdrom.com>, Chuck Robey <chuckr@glue.umd.edu>, jehamby@lightside.com, hackers@freebsd.org
Subject:   Re: FreeBSD/Alpha (was Re: COMDEX trip report) 
Message-ID:  <199612011925.LAA18010@covina.lightside.com>

next in thread | raw e-mail | index | archive | help
Mark Mayo wrote:

>On Sat, 30 Nov 1996, Jordan K. Hubbard wrote:
>> I only believe what I read in PowerPC News - You know, that
>> publication that closed its doors with a dirge in the last issue for
>> the PowerPC architecture, citing insufficient interest even among its
>> most principle sponsors (IBM / Motorola) and a failure to live up to
>> all the promised increases in performance these last 3 years.
>> 
>
>I'm curious, has the "PowerPC Platform" spec been released now?? I know it
>has been long awaited since it's suppose to finally open up the PowerPC
>clone market, and provide a standard for multiple OS machines... And can
>you buy one of these machines yet?

Yes, and soon.  IBM has their "Long Trail" motherboard, and Motorola has their
Yellowknife.  Both are available as reference implementations to clone makers,
so PPCP systems will be in full swing next year.  As for PowerPC News, I'm
inclined to agree that the PowerPC architecture was dismal as recently as
last year (when it topped out at 100MHz).  However, the 620 is back on track,
clock speeds are up (up to 225MHz), and IBM and Motorola have realistic plans
to keep the architecture competitive with Intel on performance (and with better
price/performance) in the coming years.  Unlike Intel, new CPU's from IBM and
Motorola will be available immediately on consumer machines (at lower clock speeds),
instead of coming out first for servers (i.e. Pentium Pro) and then trickling down
to consumers.  In other words, I'm more excited about PowerPC then I've ever been.

>Also, has anyone heard anything from Exponential Technology lately? Are
>they actually going to pull it off?? The last I heard they were
>predicting the first chip to appear in 2nd quarter 97 running at 300-400
>MHz, promising a 500MHz by late 97 / early 98. Are they on track? When I
>checked 2 months ago, they still hadn't put any of the design to silicon
>(or galium arsenide I suppose forthe bi-polar logic). Also, did anyone
>ever figure out what they were doing with that patent that let them share
>instructions sets in registers? The press seemed to think they were
>working on hardware emulation for the x86 (making it a PPC 615 decendant I
>guess) or the 680x0 .  They seem to be the only company offering anything
>interesting in the PowerPC line these days... I'm just not confident that
>it will happen.

Yes, they are on track!  The chip is the X704, it's available in samples, and
will be available in volume at several clock speeds, including 533MHz, next year.
The price is high ($1000/CPU) but no more than a high-end Alpha.  It is
pin-compatible with the 604, has a built-in L2 cache, and the technology
(bi-polar logic) is scalable well into the gigahertz range.  Of course it dissipates
75-85 watts of heat (!), but nothing a heat sink can't take care of, and the
PowerMac 9500 power supply is beefy enough to handle it without trouble.

The good thing about the Exponential is that it provides a high-end for the
PowerMac users in the DTP market, and pushes the price of the 604 down for
the rest of us.  And I still believe dual or quad CPU's are a more realistic
means for the highest price/performance than a single expensive CPU.  If only
more OS's recognized this and had robust SMP support...

>If the breakthroughs happen with the PowerPC (Exponential, 500 MHz,
>PowerPC Platform, BeBox) it could be a worthwhile chip to support. Until
>then, however, I think I'd like an Alpha port of FreeBSD! The Alpha is the
>fastest chip, and don't think they're about to stand aside and let some
>upstart PowerPC manufacturer to shoot past them ;-) Also, considering
>DEC's recent attempts to get into bed with Microsoft, and the pressure now
>being placed upon NT for "only being 32-bit", I think the Alpha may
>be well on it's way to becoming a competitor in the 'PC' world =)

Well, that's assuming FreeBSD/Alpha is 64-bit (reasonable, considering the Linux
port is).  As I said, the Exponential CPU looks impressive, but we'll have to wait
and see what its real-world performance is.  In the meantime, either platform is
a good choice for a FreeBSD port, but it appears the mindshare has moved
towards Alpha.  Since FreeBSD is being used more on servers, Alpha is a logical
choice when the PPro isn't enough.  Since UNIX has "lost the desktop", I guess
BeOS, MacOS, and NT Workstation, will become the big choices for PowerPC, not
AIX or Linux (or FreeBSD).

>Of course, if I could boot between MacOS, BeOS, AIX, NT, Linux,
>(?FreeBSD?) on a PowerPC Platform machine, I'd definately own one of
>those!

That's the appeal of PPCP.  In reality, most people will stick with one or two
OS's, but I'm a weirdo, and the idea of running ALL of them does appeal to me
(just as I've booted FreeBSD, Linux, Solaris, Win95, and NT from a single PC :-).
The advantage of a "universal" system lies in lower hardware prices, as the
computer becomes a commodity, much as the PC/AT standard is now.

>One more thing: wasn't Terry doing something with FreeBSD and PowerPC's ?

I think so, but I don't know if it's still on track.

>> The BeBOX is quite interesting, but not because of its dual-PowerPCs;
>> it's interesting because of the other hardware within it and its
>> operating system.
>> 
>> 					Jordan

I am 100% convinced that Be chose the right CPU with the PowerPC.  The name
of the game is not performance, but price/performance, when four commodity
CPU's can be used to outperform a single high-end Alpha, at a lower total price.
Be could have chosen Alpha, but I still believe the lack of a second-source is
going to keep volume low, and price high.  Finally, Intel is not even in the picture,
not because of the CPU, but because then people would want BeOS/x86, and the
huge difficulty of supporting all the bizarre motherboard and device driver
combinations would surely gain tiny Be a reputation worse than OS/2 as they
struggle (and fail) to support them.

-- Jake



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199612011925.LAA18010>