OT: cosmo, HTML GUi editors, graphics, games, computers, etc. (was Re: More on fP 6.0 features)

Fairlight fairlite at fairlite.com
Wed Oct 27 00:06:45 PDT 2004


On Wed, Oct 27, 2004 at 12:50:03AM -0400, Bill Vermillion, the prominent pundit,
witicized:
> 
> Well it did come with the system.  At prices like that you do want
> productivity tools included.

Well...yeahhhhh...  (As in, duhhh...)  *grin*  "Thanks for choosing SGI
today.  Your total is $17000.  Would you like productivity tools with that?
Thank you, please drive ahead."  *giggle*  I can't help it, man...it just
kinda comes out at this point.  I -really- will make a concerted effort to
get >4hrs sleep tonight.

> And I forgot what the name of the other program was that you could
> use.  It was amazing at that time.  You would see a scene and you
> could pan 360 degrees around it - using photographs made with a
> camera that pivoted 360 degrees to expose the pictures - and then
> stitch them together.
> 
> This was when SGI was riding high with film work like Jurassic
> Park.  I think I sent Jay a copy of the 8MB .tiff file they had
> that was part of their poster that said "Building a better
> dinosaur".

Not Blender, was it?  That was a Neo-Geo kit that was originally only
available on SGI's, and then later ported to linux.  That was some -really-
cool software--if you could figure it out.  The linux port had -no-
documenation at the time I tried it.  What I managed to figure out was
really cool.  But SGI was the original native platform for Blender.

> That's sort of like not having good luck with a system running
> an Intel 8008 and blaming all of Intel - missing such things
> as the Xeons with 4MB of L1 cache.

Well, there is that.  But at the time, it was state of the art, and
supposedly superior.  I have my doubts, as my PSX performed better at
32bit.

> I forget which chip was to be the end of the line - 10000 or 12000,
> but then Intel got so far behind with their 64x64 chip [the failure
> to deliver was one reason Project Monterey fell apart and SCO,
> IBM and I think HP parted company on that].  So they came out with
> the 14000.  It and the Alpha were the only true 64x64 chips
> available.  Real screamers.

Isn't there an IA64?

> > Maybe they overtaxed it in some games, but I'm talking -first-
> > generation games here--and usually you don't start stressing
> > the console's limitations until about the fourth year and
> > generation of games.
> 
> What model number MIPs.  They had at least as many models over the
> years as Intel did.

More like Motorola from the looks of it.  Here's what I dug up:

     * CPU: MIPS R4300i, 93.75MHz, 64-bit, 24KB L1, 125 MIPS, 250 MB/sec Bus
     * Graphics: SGI RCP, 62.5MHz, 100 MFLOPS, 150K Polygons/Sec, 32-bit 
       Color, 500 MB/sec Bus
     * Sound: SGI RCP, 64 2D Voices, ADPCM, 500 MB/sec Bus
     * Data: 4MB (500 MB/s), Cartridge (32MB), Expansion 4MB RAM

And strangely, though the chip is listed as 64bit even in this emulator
spec summary, they say that it really wasn't 64-bit, but rather an
advertising spin on two 32bit processors.  Yet, the chip is listed as
64bit.  I dunno, I can find multiple sources that say it was a custom (all
Nintendo's chips are custom versions of stock chips--originally from
Motorola, this time from MIPS, and the latest is a custom PowerPC) MIPS
4300, and there's even a PDF on one site (didn't bother downloading
it--half the architecture low-level stuff will breeze right over my head,
as I don't do IC design).  It's the same kind of confusion that has
Nintendo's own manuals claiming the -original- GameBoy was a custom 65c02,
but I've talked to people that took them apart and claimed they were
z80+'s.  But every source I just looked at said MIPS 4300 (either 'i' or
custom).

> I didn't have any problems.   What version of IRIX.  When the 5.3
> ?? came out and all the COFF files went away for ELF, the install
> was the nicest I've ever run.  It checked everything.  And the
> application installs were the best I've ever seen. Supreme
> dependancy checking for everything with choices all along the way.

The version that would have been around in about '94.  *looks overly
helpful*  I never kept up with it that much, and the only bits I remember
were pure compiler hell.

> You had a choice.  I ran in console mode a great deal of the time.
> Working with vi with READABLE lines of 200+ character lines all
> on the screen was nice.  I worked in text mode most of the time.

Interesting.

> I think I mentioned that too.  A friend of mine who retired with a
> big wad of money after 4 years of 24x7 on call when Verio bought
> out the ISP he was head tech at once told me that he'd never
> put an SGI or and SCO system on the 'net.

Yup, those were the two.  I'd trust SCO systems on the net--either behind a
firewall or if they were decently secured by one of the few people I trust
to do the job correctly.  The folks that did the SGI's I'm thinking of
across town were not the types that inspired confidence, based on some of
the practices I heard went on.

> So one day he sat down, disassembled the kernel, traced the code,
> and found the problem.  AT&T had screwed up in re-scanning for free
> blocks, and in one spot the system thought it had scanned from the
> front and it had not.  So he developed a binary patch that all the
> Microport users had, but it took almost a year before AT&T
> put the fix into the releases and the rest of the Unix world
> didn't have to worry about that but.

I remember this story.  Sometimes you reiterate one, but they're always
interesting nonetheless.  I find that one incredible.  That's the -real-
sign of a bug that won't let you get to sleep, lemme tellya.

> So if/when I get a really bad problem I can give him a call.

Kinda like when I call you.  :)

> So I believed him when he said he'd never put those machines on the
> .net.  I was running a pair of Challeng servers.

Gack!  "Turn them off!  Turn them OFF!  NOW!!"  I can see it.  :)

> I think they were 400MHz with ECC memory - probably 500MB.

Pricey, considering they're basically all custom.  Ouch.

> That machaine was running a 120MHz Pentium with 128MB of RAM.
> It ran faster than the SGI.   But the FreeBSD's of that era
> were exceptionally lean.

You know, I have that problem with linux nowadays.  1.2, 2.0, 2.2...they
were all fast.  They got to 2.4, and I went from being able to boot linux
three times in the time it would take to boot Win95 -once-...to the point
where linux actually takes 20-30 seconds longer to load on the same exact
P166--not just identical, the -same- system.  I dunno what they did to the
2.4 kernels, but I'm not pleased with it in the least.  I hope 2.6 fares
better, and they finally worked out the memory management.

The days when you could claim to run linux from a single floppy on a 386
with 4MB of RAM are long gone, IMHO.  I don't think it's technically
possible.  Used to be, and it was handy.  But there are tradeoffs.  I
wouldn't go back all that far, but I'd -like- to see the speed of 2.2
brought back.  I'm almost certain the decrease resides in the memory
management.  It's the most likely suspect.  Unless ReiserFS is slower than
ext2, but I wouldn't expect it to be.  Hmmm.  Well, I'm running ancient
crud here anyway, except my wife's machine.  On spanking new machines,
nobody notices one way or another--but I think they'd scream even faster
on those dual and quad Xeon 2.4GHz hyperthreaded systems if they had a 2.2
kernel. :)

Linux just isn't near as lean anymore either.  FreeBSD is not alone in that
respect, I'm afraid.

Ah, progress...  *wistful sigh*

mark->
-- 
Bring the web-enabling power of OneGate to -your- filePro applications today!

Try the live filePro-based, OneGate-enabled demo at the following URL:
               http://www2.onnik.com/~fairlite/flfssindex.html


More information about the Filepro-list mailing list