Windows NT & StrongARM
Tower Electronics Ltd
Dear NewsNet Reader,
I have received permission to post this on
the Acorn Newsgroups. For further info please E mail
the author directly. ITEM START-----------------------------
Received: from pool025.Max1.Miami.FL.DYNIP.ALTER.NET (pool025.Max1.Miami.FL=
.DYNIP.ALTER.NET [153.37.90.25]) by germany.it.earthlink.net (8.7.5/8.7.3) =
with SMTP id IAA28834 for <to...@enterprise.net>; Thu, 6 Jun 1996 08:07:17 =
-0700 (PDT)
X-Authentication-Warning: germany.it.earthlink.net: Host pool025.Max1.Miami=
.FL.DYNIP.ALTER.NET [153.37.90.25] didn't use HELO protocol
Message-ID: <31B6F4...@earthlink.net>
Date: Thu, 06 Jun 1996 10:08:03 -0500
From: Frode Wells <frode...@earthlink.net>
Reply-To: frode...@earthlink.net
X-Mailer: Mozilla 3.0b4Gold (Win95; I)
MIME-Version: 1.0
To: to...@enterprise.net
Subject: So You Think Acorn is Not Industry Standard?
Content-Type: text/plain; charset=3Dus-ascii
Content-Transfer-Encoding: 7bit
I went to a corporate briefing hosted by Microsoft a few months ago. I
asked both Microsoft & Digital Co. if Windows NT will support StrongARM.
None of them could give me a firm no or yes. The answer was maybe. If
the case is that Windows NT will support StrongARM this would be bad
news for PentiumPro & Intel Co.
Just imagine a RiscPC with two StrongARMs running Windows NT & RiscOS.
Running Windows NT with a StrongARM as a graphics accelerator - it should
be rock steady.
Let there be no question about it I prefer RiscOS & I am very happy with my=
RiscPC. Most people in USA do not seem to understand that RiscOS is a
better platfrom than Windows.
P.S.! You could know that Intel Co. are working hard on their next
generation chips, RISC chips. These chips will replace the PentiumPro
sometime in 1998. They are counting on the PentiumPro & their multimedia
chips in 1997. If you buy a PentiumPro computer you would not get much
for it when you sell it, unlike Acorn computers they hold their value
much longer.
\|/
(. .)
----------------------------------------------------o00-(_)-00o-----
Frode M. Wells=09=09=09 Phone: +1 (305) 861-2931
835 82nd Street #3=09=09 Email: frode...@earthlink.net
Miami Beach, FL 33141-1370, USA
--------------------------ITEM ENDS----------------------------------------=
-
Tom Waller. Tower Electronics Ltd. The Lewes, Main Street, Fyvie
Turriff Aberdeenshire Scotland UK AB53 9BY
Tel UK (44) 01651 891069 Fax UK (44) 01651 891653
http://www.enterprise.net/tower-risc
Steve Jelfs
In article <ant10174...@tower.enterprise.net>
Tower Electronics Ltd <to...@enterprise.net> wrote:
> Dear NewsNet Reader,
[lots of header snipped]
> I went to a corporate briefing hosted by Microsoft a few months ago. I
> asked both Microsoft & Digital Co. if Windows NT will support StrongARM.
> None of them could give me a firm no or yes. The answer was maybe. If
> the case is that Windows NT will support StrongARM this would be bad
> news for PentiumPro & Intel Co.
>
Could it be that they had never heard of StrongArm and therefore could not possibly be in a position to confirm or deny? Sounds like straw clutching to me.
[lots mor snipped]
--
Steve Jelfs
st...@jelfs.demon.co.uk
sd-j...@wpg.uwe.ac.uk
sje...@brookes.ac.uk
... Hailing frequencies open Mr. Worf. - Hi, this is Chris Evans on 1 FM.
Ian Griffiths
Is this a troll?
Tower Electronics Ltd wrote:
>
> If the case is that Windows NT will support StrongARM this
> would be bad news for PentiumPro & Intel Co.
Why exactly? There are already 3 high performance RISC
architectures around which will run NT.
> Just imagine a RiscPC with two StrongARMs running Windows NT &
> RiscOS. Running Windows NT with a StrongARM as a graphics
> accelerator - it should be rock steady.
Now imagine a machine with two PowerPCs running Windows NT, and
a Matrox Millenium graphics accelerator.
My money's on the PPC machine when it comes to speed.
Hell a machine with two P133s and a Matrox Millenium is pretty
darn fast.
--
Ian Griffiths
David Thornton
In article <35454...@jelfs.demon.co.uk>, Steve Jelfs
<st...@jelfs.demon.co.uk> writes
>Could it be that they had never heard of StrongArm and therefore could not
>possibly be in a position to confirm or deny? Sounds like straw clutching to
>me.
I emailed the MS Windows NT project Manager a few weeks ago to ask if
Windows NT would be ported to the StrongARM. He stated that "it doesn't
seem likely."
Windows NT runs best on x86 processors if you are mostly interested in
software compatibility. There are many apps which won't run under the
PowerPC version of NT.
--
David Thornton
da...@modcon.demon.co.uk
Alistair M Cockburn
> Tower Electronics Ltd wrote:
> >
> > If the case is that Windows NT will support StrongARM this
> > would be bad news for PentiumPro & Intel Co.
> > RiscOS. Running Windows NT with a StrongARM as a graphics
> > accelerator - it should be rock steady.
Why stop at 2? The Hydra can do many more :)
--
"A radical is a man with both feet planted firmly in the air"
Thomas Down
Alistair M Cockburn (cock...@argonet.co.uk) wrote:
: > Tower Electronics Ltd wrote:
: > >
: > > If the case is that Windows NT will support StrongARM this
: > > would be bad news for PentiumPro & Intel Co.
: > > Just imagine a RiscPC with two StrongARMs running Windows NT &
: > > RiscOS. Running Windows NT with a StrongARM as a graphics
: > > accelerator - it should be rock steady.
: Why stop at 2? The Hydra can do many more :)
Because Windows NT multi processor supports is very non-scalable, and
the benefits of having more than 2 processors diminish very rapidly..
-- Thomas
"I am a Keeper, and responsible only to my own conscience."
- Elorie, The Bloody Sun
Tower Electronics Ltd
In article <31BD71...@dev.madge.com>, Ian Griffiths
NO TOWER ELECTRONICS DID NOT WRITE SIMPLY RELAYED THE original E mail
from america::::
> > If the case is that Windows NT will support StrongARM this
> > would be bad news for PentiumPro & Intel Co.
>
> Why exactly? There are already 3 high performance RISC
> architectures around which will run NT.
AND HAVE YOU LOOKED AT THE PRICE TAG - go on then purchase your
Sun Sparc or HP/IBM Risc and see how much change you get out of
60,000 ukp! A strongARM based Risc is faster than an Ultra Sparc
and over 65 mips faster than a 90Mhz clock speed Pentium
(These are direct quotes from July edition of the Computer Shopper
magazine - not that well known for its glowing reviews of Acorn
hardware......
> > Just imagine a RiscPC with two StrongARMs running Windows NT &
> > RiscOS. Running Windows NT with a StrongARM as a graphics
> > accelerator - it should be rock steady.
>
> Now imagine a machine with two PowerPCs running Windows NT, and
> a Matrox Millenium graphics accelerator.
Risc PC's do not require a graphics accelerator to display in up
to 32 thousand screen colours at 1024 by 768 pixels when fitted
with 2Mb Vram and most rekon that the next generation of ART
design will improve on this again.
> My money's on the PPC machine when it comes to speed.
You are wrong - see my commercial Web pages for direct quotes from
various sources who now see the PPC ref platform as pretty jaded.
> Hell a machine with two P133s and a Matrox Millenium is pretty
> darn fast.
TRUE.... if you have any to give away for "research and review"...
> Ian Griffiths
Ian Griffiths
Tower Electronics Ltd wrote:
>
> Ian Griffiths <mailto:igri...@dev.madge.com> wrote:
> >
> > Is this a troll?
> >
> > Tower Electronics Ltd wrote:
>
> NO TOWER ELECTRONICS DID NOT WRITE SIMPLY RELAYED THE original
I somehow suspect that America didn't write that. Maybe someone
in America did.
But anyway, I will correct it:
> > Tower Electronics posted, but didn't provide any useful
> > attribution for this:
:-)
> > > If the case is that Windows NT will support StrongARM this
> > > would be bad news for PentiumPro & Intel Co.
> >
> > Why exactly? There are already 3 high performance RISC
> > architectures around which will run NT.
>
> AND HAVE YOU LOOKED AT THE PRICE TAG - go on then purchase
> your Sun Sparc or HP/IBM Risc
Well the Sun won't run NT of course, and neither will the HP. I
didn't think that PowerPC machines were that expensive.
PowerMACs certainly aren't.
> and see how much change you get out of 60,000 ukp!
For a PowerMAC? Quite a lot actually.
And the next generation PowerMACs will be able to run NT.
I don't have any appropriate price lists to hand. Can you tell
me how much a PPC-based PC costs? I was guessing at sub-2000
UKP. I'm certain it's not 65k.
> A strongARM based Risc is faster than an Ultra Sparc
> and over 65 mips faster than a 90Mhz clock speed Pentium
And by the time a StrongARM version of NT was available? (Not
that that will happen.) The top of the range Pentium is faster
than a P90 by the way. If you are going to make comparisons,
remember that the StrongARM is the very fastest ARM chip around,
and you can't actually go and buy machines with them in yet. So
making comparisons with a slow incarnation of Intel's previous
line of chips is not very meaningful.
> (These are direct quotes from July edition of the Computer
> Shopper magazine - not that well known for its glowing reviews
> of Acorn hardware......
Surprising. Did it contain any useful information too? Have
you edited it to your purposes just as you've done with the
Usenet subjects on your home page?
> > Now imagine a machine with two PowerPCs running Windows NT,
> > and a Matrox Millenium graphics accelerator.
>
> Risc PC's do not require a graphics accelerator to display in
> up to 32 thousand screen colours at 1024 by 768 pixels when
> fitted with 2Mb Vram
No you're right there. However, my Matrox millenium with 2MB of
VRAM will do 1024x768 in 65536 colours too. How do you make
your RiscPC do that? I can upgrade my Matrox Millenium to run
at 1600x1200 in 16 million colours. How do you make your RiscPC
do that?
> and most rekon that the next generation of ART design will
> improve on this again.
Do they? I'm sure you'd like to think that. ART have suggested
that a 4MB VRAM system is a possibility. That will be an
improvement, but it will still be behind today's possibilities
on PCs. Furthermore, the lack of a graphics accelerator will
start to be a problem when more VRAM is available, because the
processor will be having to work twice as hard, and since for
graphics it's mostly bus limited, the StrongARM won't make
things any better.
> > My money's on the PPC machine when it comes to speed.
>
> You are wrong - see my commercial Web pages for direct quotes
> from various sources who now see the PPC ref platform as
> pretty jaded.
The only reference to PPC machines I could find was buried in
the Tower Electronics News. There are no quotes from any
sources so far as I can tell. Would you care to provide a
specific URL so we can read these?
It's called PPCP now not CHRP by the way.
Incidentally your 'these pages look best with Antialiasing'
picture is a little unfair - you've drawn the un-antialiased
text at half the resolution, which is misleading.
> > Hell a machine with two P133s and a Matrox Millenium is
> > pretty darn fast.
>
> TRUE.... if you have any to give away for "research and
> review"...
I only have the one I'm afraid, and I'm not exactly inclined to
part with it at the moment.
--
Ian Griffiths
Greg Hennessy
In article <ant13060...@tower.enterprise.net>,
Tower Electronics Ltd <to...@enterprise.net> writes:
>>
>> Why exactly? There are already 3 high performance RISC
>> architectures around which will run NT.
>
>AND HAVE YOU LOOKED AT THE PRICE TAG - go on then purchase your
>60,000 ukp!
Yup and 60k would buy a very nice machine with quite a lot change
left over. The HP 712/80 that this is being typed on was substantially
less than that, it having 128mb ram and 4 GB of disk space.
> A strongARM based Risc is faster than an Ultra Sparc
Hmmm.... I would like to see how it compares to an Ultra Enterprise 170, I
have yet to see any Arm based box running Oracle 7 or the Netscape WWW
servers, or for that matter any mainstream commercial software. Kinda
reduces it's utility somewhat for practical work, And I wont even mention
useful stuff like hardware FP.
>and over 65 mips faster than a 90Mhz clock speed Pentium
More 'Meaningless Indication of Processor Speed', It might as well be
65 bips faster for use they would be considering the lack of software.
>
> Risc PC's do not require a graphics accelerator to display in up
>to 32 thousand screen colours at 1024 by 768 pixels when fitted
>with 2Mb Vram and most rekon that the next generation of ART
>design will improve on this again.
>
Hmm I run multi-threaded, PMT O/S's, I dont want my CPU tied up
re-drawing windows or blitting graphics around the place.
>> My money's on the PPC machine when it comes to speed.
>
>You are wrong - see my commercial Web pages for direct quotes from
>various sources who now see the PPC ref platform as pretty jaded.
>
Hmm... I have seen one or two Aix boxes with a PPC onboard and
they weren't what one would call jaded.
greg
--
Greg Hennessy
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|~~~~~~~~~~~~~~~~~~~~~~~~~~~~
ZZ Top, They Really Can't Be Beat |ghen...@www.lhr-sys.dhl.com
Webmasters do it with <HTML> |cmk...@cix.compulink.co.uk
Nunzz!! Aaargh !!! |gr...@cmkrnl.demon.co.uk
Tony Houghton
In article <Dsy1s...@systems.DHL.COM>
ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
> Hmmm.... I would like to see how it compares to an Ultra Enterprise 170, I
> have yet to see any Arm based box running Oracle 7 or the Netscape WWW
> servers, or for that matter any mainstream commercial software.
If you want to see Netscape running on an ARM try looking at the picture
of RiscBSD in Risc User. Or wait until Netscape port it to the NC.
> Kinda
> reduces it's utility somewhat for practical work,
Not when there's equally practical software available for the majority
of tasks. In fact more practical if you consider it isn't overburdened
with never-used features, doesn't have to fight through a heavyweight
OS, and is affordable.
> And I wont even mention
> useful stuff like hardware FP.
You just did.
> >and over 65 mips faster than a 90Mhz clock speed Pentium
>
> More 'Meaningless Indication of Processor Speed', It might as well be
> 65 bips faster for use they would be considering the lack of software.
Sigh. How long are people going to keep on spouting that ridiculous
mantra? All right, maybe someone's inaccurate over-enthusiasm got you
wound up, but why wind everyone up with more inaccurate irrelevancies?
> > Risc PC's do not require a graphics accelerator to display in up
> >to 32 thousand screen colours at 1024 by 768 pixels when fitted
> >with 2Mb Vram and most rekon that the next generation of ART
> >design will improve on this again.
> >
> Hmm I run multi-threaded, PMT O/S's, I dont want my CPU tied up
> re-drawing windows or blitting graphics around the place.
Some people occasionally concentrate on one graphics-intensive task.
They don't want to sit twiddling their thumbs while the CPU waits for
the graphics accelerator's far slower processor to draw the screen.
--
http://www.tcp.co.uk/~tonyh/
for WinEd, Bombz and miscellaneous utilities for RISC OS
Tony Houghton
In article <31C012...@dev.madge.com>
Ian Griffiths <igri...@dev.madge.com> wrote:
> No you're right there. However, my Matrox millenium with 2MB of
> VRAM will do 1024x768 in 65536 colours too. How do you make
> your RiscPC do that? I can upgrade my Matrox Millenium to run
> at 1600x1200 in 16 million colours. How do you make your RiscPC
> do that?
How are you going to make your already very responsive PC 4-5x faster
for under 300UKP at the end of summer? ;)
> > and most rekon that the next generation of ART design will
> > improve on this again.
>
> Do they? I'm sure you'd like to think that. ART have suggested
> that a 4MB VRAM system is a possibility. That will be an
> improvement, but it will still be behind today's possibilities
> on PCs.
What makes you think that 4MB VRAM is not an improvement on 2MB?
> Furthermore, the lack of a graphics accelerator will
> start to be a problem when more VRAM is available, because the
> processor will be having to work twice as hard, and since for
> graphics it's mostly bus limited, the StrongARM won't make
> things any better.
It will in fact. My money's on the StrongARM being faster than a Matrox
Millenium. I don't think that just because it's a dedicated graphics
card it means it can magically handle 4-8MB VRAM just as fast as 1-2MB.
Greg Hennessy
In article <c31c...@tonyh.tcp.co.uk>,
Tony Houghton <to...@tcp.co.uk> writes:
>In article <Dsy1s...@systems.DHL.COM>
> ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
>
>If you want to see Netscape running on an ARM try looking at the picture
>of RiscBSD in Risc User. Or wait until Netscape port it to the NC.
>
Note I did say 'WWW servers', not the browser. Netscape running over X does
not count here. I wasn't the one to make a comparison with the ultra sparc
but seeing that someone is trying to make an apples to apples comparison,
Any current or future ARM based system is likely to be found wanting as
commercial applications or DBMS server platform. NT for Strongarm is
just not going to happen.
>> Kinda
>> reduces it's utility somewhat for practical work,
>
>Not when there's equally practical software available for the majority
>of tasks. In fact more practical if you consider it isn't overburdened
>with never-used features, doesn't have to fight through a heavyweight
>OS, and is affordable.
Hmmm... I manage approximately 2000 www users who make active use of
ms-office as helper apps, I dont see the Arm version anywhere. I provide
a searchable WWW interface into a document repository containing > 1gb
of .doc, .xls, .ppt files, I use netscape's commerce and proxy servers,
please tell me how your 'practical software' could plug into this
environment either as a client or a server and allow my users to work as
they do at the moment ?
Whats this about a heavyweight OS ? I dont consider basic OS features like
PMT, VM and Threading to be heavyweight. Just because the developers of
RISC-OS have in their infinite wisdom declined to put these features in
their OS does not reduce their utility on other platforms. As a matter of
fact IMHO it just goes to show that whilst TPTB at acorn have sat on
their collective arses for the past 6/7 years O/S wise the world has
moved on and their product alas is getting long in the tooth.
>
>> And I wont even mention
>> useful stuff like hardware FP.
>
>You just did.
>
:-) Vindication is just so sweet.
>> >and over 65 mips faster than a 90Mhz clock speed Pentium
>>
>> More 'Meaningless Indication of Processor Speed', It might as well be
>> 65 bips faster for use they would be considering the lack of software.
>
>Sigh. How long are people going to keep on spouting that ridiculous
>mantra? All right, maybe someone's inaccurate over-enthusiasm got you
>wound up, but why wind everyone up with more inaccurate irrelevancies?
>
See my point above for further details.
>> > Risc PC's do not require a graphics accelerator to display in up
>> >to 32 thousand screen colours at 1024 by 768 pixels when fitted
>> >with 2Mb Vram and most rekon that the next generation of ART
>> >design will improve on this again.
>> >
>> Hmm I run multi-threaded, PMT O/S's, I dont want my CPU tied up
>> re-drawing windows or blitting graphics around the place.
>
>Some people occasionally concentrate on one graphics-intensive task.
>They don't want to sit twiddling their thumbs while the CPU waits for
>the graphics accelerator's far slower processor to draw the screen.
>
ROFL. You are being slightly naive here. ;-). Check out a Matrox
Millenium sometime and then come back to me with the above assertion.
Shucks even a 70 quid S3 Trio based board will do the job quite nicely.
Some choice have the CPU blit a window over a seriously bus limited
architecture or just send a MoveRect primitive direct to the accelerator's
command queue and return immediately to do some more work. Also most modern
graphic accelerators have useful items like save under and font caches
builtin as default with no additional overhead to either the OS or the CPU.
Ian Griffiths
Tony Houghton wrote:
>
> Ian Griffiths <igri...@dev.madge.com> wrote:
>
> > No you're right there. However, my Matrox millenium with
> > 2MB of VRAM will do 1024x768 in 65536 colours too. How do
> > you make your RiscPC do that? I can upgrade my Matrox
> > Millenium to run at 1600x1200 in 16 million colours. How do
> > you make your RiscPC do that?
>
> How are you going to make your already very responsive PC 4-5x
> faster for under 300UKP at the end of summer? ;)
It's already faster than all my Acorn machines. :-)
> > > and most rekon that the next generation of ART design will
> > > improve on this again.
> >
> > Do they? I'm sure you'd like to think that. ART have
> > suggested that a 4MB VRAM system is a possibility. That
> > will be an improvement, but it will still be behind today's
> > possibilities on PCs.
>
> What makes you think that 4MB VRAM is not an improvement on
> 2MB?
I didn't say that it wasn't. What I was arguing against was the
suggestion that graphics cards are necessarily a bad thing.
That has rather got lost in the remaining quotes here, but that
was my point.
Yes 4MB is better than 2. It's not as good as 8. On a 32 bit
system, it will be slower with 4MB than a PC 4MB graphics card.
> > Furthermore, the lack of a graphics accelerator will
> > start to be a problem when more VRAM is available, because
> > the processor will be having to work twice as hard, and
> > since for graphics it's mostly bus limited, the StrongARM
> > won't make things any better.
>
> It will in fact. My money's on the StrongARM being faster than
> a Matrox Millenium. I don't think that just because it's a
> dedicated graphics card it means it can magically handle 4-8MB
> VRAM just as fast as 1-2MB.
No, but the fact that it's bus is twice as wide means it can
handle 8MB just as fast as a 32 bit 4MB system. No magic
involved.
My point was that for the majority of graphics output, the bus
is saturated with existing ARMs. A StrongARM will just be
stalled for more clock cycles, it won't actually go any faster,
except for stuff which really is slow enough on current machines
to be limited by processor speed. Artworks (once it works) will
probably go faster on things like graduated fills. Dragging
windows around will not get very much faster (except perhaps
when you're dragging it over complex windows) and indeed will
get a lot slower when 4MB VRAM machines turn up, unless Acorn do
something to make the bus faster.
There is no magic involved, it's just higher bandwidth.
--
Ian Griffiths
Ian Lynch
Ian Griffiths <igri...@dev.madge.com> wrote:
> Dragging windows around will not get very much faster (except perhaps
> when you're dragging it over complex windows)
Why do you want to drag windows faster? I pull on the mouse and the
window moves to where I want it. Ok it ain't rock steady on RPC 600
but so what? I am not normally trying to read the thing while its
moving. Now screen re-draws, that is something which would be better a
bit faster and rendering some graphics. I understood from the tests
using the Artworks viewer that these would be a lot faster. Are you
saying that they aren't so I'd be better off saving my money?
--
Ian
Tony Houghton
In article <Dsz92...@systems.DHL.COM>
ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
> Note I did say 'WWW servers', not the browser. Netscape running over X does
> not count here. I wasn't the one to make a comparison with the ultra sparc
> but seeing that someone is trying to make an apples to apples comparison,
> Any current or future ARM based system is likely to be found wanting as
> commercial applications or DBMS server platform. NT for Strongarm is
> just not going to happen.
Point taken, but there is an HTML server for RISC OS. How it compares
with Netscape I don't know, but all the web pages I've ever seen merely
consist of a server chucking data at me when asked and occasionally
running a script. There's no reason the RISC OS version couldn't do
that.
> Hmmm... I manage approximately 2000 www users who make active use of
> ms-office as helper apps, I dont see the Arm version anywhere. I provide
> a searchable WWW interface into a document repository containing > 1gb
> of .doc, .xls, .ppt files, I use netscape's commerce and proxy servers,
> please tell me how your 'practical software' could plug into this
> environment either as a client or a server and allow my users to work as
> they do at the moment ?
Computers are becoming more data centred than application centred. If
you're starting with a WWW interface and making it into something that
depends on proprietary 'standards' you're headed in the wrong
direction. Maybe you had no choice at the time.
> Whats this about a heavyweight OS ? I dont consider basic OS features like
> PMT, VM and Threading to be heavyweight. Just because the developers of
> RISC-OS have in their infinite wisdom declined to put these features in
> their OS does not reduce their utility on other platforms. As a matter of
> fact IMHO it just goes to show that whilst TPTB at acorn have sat on
> their collective arses for the past 6/7 years O/S wise the world has
> moved on and their product alas is getting long in the tooth.
Undoubtedly RISC OS would be better off with PMT, VM and Threading. But
for some reason every OS that has these features suffers from bloat. VM
is just a necessity to get them working at all instead of a useful
enhancement for working with large data, which is now supported on most
RISC OS applications that need it. The lack of modern kernel features
is mainly a problem to programmers - I am a programmer, yet I still
forgive it - the user still gets a working environment about as good as
any for many people's needs.
> >Some people occasionally concentrate on one graphics-intensive task.
> >They don't want to sit twiddling their thumbs while the CPU waits for
> >the graphics accelerator's far slower processor to draw the screen.
> >
>
> ROFL. You are being slightly naive here. ;-). Check out a Matrox
> Millenium sometime and then come back to me with the above assertion.
> Shucks even a 70 quid S3 Trio based board will do the job quite nicely.
My point was that some work requires you to be able to see the result
of the previous step before you can continue, so the processor being
available while a redraw is going on isn't much use to you. If a Matrox
Millenium could make redraws instantaneous, then so could a StrongARM
or good Pentium. The difference in performance wouldn't be nearly as
noticeable as the difference in cost between a dedicated high-
performance graphics card and relying on a (faster) general-purpose
processor that you've already got.
The Risc PC was never marketed as a professional graphics workstation,
so why criticise it for not being as capable as a PC with an expensive
add-on to make it into one? I don't think a family MPC from a box-
shifter is likely to have a graphics card that will significantly
outperform a Risc PC.
You're obviously no ordinary user don't forget. Most ordinary users
don't know that they can multi-task at all, so whether it's pre-emptive
or not doesn't matter to them. Neither do they need better graphics
capabilities than what's available with 2MB VRAM and limited bandwidth.
The Risc PC's more logical interface would suit them better, and if
more people used it it would generate a market for people who want,
"What everybody else has got, only better," and push development. It
would also make the job of people like you (if I've got the right
impression of what you do) easier. Or perhaps you don't want it to be
so easy that you're not needed. >:->
Phillip Temple
In message <31C012...@dev.madge.com> Ian Griffiths wrote:
> But anyway, I will correct it:
>
>
> > > Tower Electronics posted, but didn't provide any useful
> > > attribution for this:
>
> :-)
An entire set of SMTP headers was included, and the original
author's full signature. What else did you want?
> No you're right there. However, my Matrox millenium with 2MB of
> VRAM will do 1024x768 in 65536 colours too. How do you make
> your RiscPC do that? I can upgrade my Matrox Millenium to run
> at 1600x1200 in 16 million colours. How do you make your RiscPC
> do that?
How do you make your RPC do that? It is one of the Acorn pre-defined
modes. I always use 1024x768 with 65536 colours. It takes 1536k.
You can upgrade to 1600x1200 16M? I make that 1600*1200*4 bytes,
which is 7.5Mb of VRAM! Is this upgrade quite expensive?
> > and most rekon that the next generation of ART design will
> > improve on this again.
>
> Do they? I'm sure you'd like to think that. ART have suggested
> that a 4MB VRAM system is a possibility. That will be an
> improvement, but it will still be behind today's possibilities
> on PCs. Furthermore, the lack of a graphics accelerator will
> start to be a problem when more VRAM is available, because the
> processor will be having to work twice as hard, and since for
> graphics it's mostly bus limited, the StrongARM won't make
> things any better.
ART can put in 4Mb easily enough if needed, using 2 interleaved
banks. Whether the VIDC will be able to take advantage of this
is another question...
--
Phillip Temple
Tony Houghton
In article <31C17C...@dev.madge.com>
Ian Griffiths <igri...@dev.madge.com> wrote:
> Tony Houghton wrote:
> >
> > How are you going to make your already very responsive PC 4-5x
> > faster for under 300UKP at the end of summer? ;)
>
> It's already faster than all my Acorn machines. :-)
Evasion!
> Yes 4MB is better than 2. It's not as good as 8. On a 32 bit
> system, it will be slower with 4MB than a PC 4MB graphics card.
Not necessarily, it depends on far too many variables to make that
generalisation.
> My point was that for the majority of graphics output, the bus
> is saturated with existing ARMs. A StrongARM will just be
> stalled for more clock cycles, it won't actually go any faster,
> except for stuff which really is slow enough on current machines
> to be limited by processor speed.
All the evidence proves you wrong.
> Artworks (once it works) will
> probably go faster on things like graduated fills. Dragging
> windows around will not get very much faster (except perhaps
> when you're dragging it over complex windows) and indeed will
> get a lot slower when 4MB VRAM machines turn up, unless Acorn do
> something to make the bus faster.
They said they would. There doesn't seem much point in making the bus
faster than the RAM, but they are at least going to use EDO in future.
Richard Travers
Phillip Temple (ho...@infopark.demon.co.uk) wrote:
> How do you make your RPC do that? It is one of the Acorn pre-defined
> modes. I always use 1024x768 with 65536 colours. It takes 1536k.
Hey, have I missed something here? My RPC will run in 32K colours and
16M colours, but I'm damned if I can get it to run in 64K colours.
Not that I actually want it to, of course ;-)
--
---------------------------------------------------------------------------
| Richard Travers |*ZFC*| 01736 | Obfuscation |
| ri...@argonet.co.uk |**B**| 757941 | to order |
---------------------------------------------------------------------------
Stuart Halliday
Richard Travers (ri...@argonet.co.uk) wrote the following...
> Phillip Temple (ho...@infopark.demon.co.uk) wrote:
>
> > How do you make your RPC do that? It is one of the Acorn pre-defined
> > modes. I always use 1024x768 with 65536 colours. It takes 1536k.
>
> Hey, have I missed something here? My RPC will run in 32K colours and
> 16M colours, but I'm damned if I can get it to run in 64K colours.
>
> Not that I actually want it to, of course ;-)
The RiscPC/A7000 doesn't have a 64,000 colour mode. He must have been mistaken
or he was talking about a PC.
BTW, doesn't a PC video card running a 64,000 colour mode only have a 18-bit
Palette? I remember a PC programmer complaining about this when he had to
support 64K colour modes in a DOS program.
Oops! Dropped off newgroup topic! ;-)
--
Stuart Halliday - Web Master of the
___ ___ _ __ ___ _ _
/ _ \ __ ___ _ _ _ _ / __| _| |__ ___ _ \ \ / (_) | |__ _ __ _ ___
| _ / _/ _ \ '_| ' \ | (_| || | '_ \/ -_) '_\ V /| | | / _` / _` / -_)
|_| |_\__\___/_| |_||_| \___\_, |_.__/\___|_| \_/ |_|_|_\__,_\__, \___|
|__/ |___/
http://www.cybervillage.co.uk/acorn/
ftp://quantum:qua...@ftp.cybervillage.co.uk/pub
Peter Smith
In message <internews...@argonet.co.uk> Richard Travers wrote:
> Phillip Temple (ho...@infopark.demon.co.uk) wrote:
>
> > How do you make your RPC do that? It is one of the Acorn pre-defined
> > modes. I always use 1024x768 with 65536 colours. It takes 1536k.
>
> Hey, have I missed something here? My RPC will run in 32K colours and
> 16M colours, but I'm damned if I can get it to run in 64K colours.
>
> Not that I actually want it to, of course ;-)
Seeing as how the RPC uses 16bits for a 32k colour screenmode, strictly
speaking there _are_ 64k colours. There are 5 bits each of red, green and
blue, and as far as I understand the remaining bit used as a mask (alpha
channel?).
Similarly 24bit colour takes up 1876k. This is because there are 8
bits each of red, green and blue, and an 8 bit alpha channel. Well handy.
(Alledgedly :-)
--
___ _ ___ _ _ _
| _ \___| |_ ___ _ _ / __|_ __ (_) |_| |_
| _/ -_) _/ -_) '_| \__ \ ' \| | _| ' \
|_| \___|\__\___|_| |___/_|_|_|_|\__|_||_|
Phillip Temple
In message <internews...@argonet.co.uk> Richard Travers wrote:
> Phillip Temple (ho...@infopark.demon.co.uk) wrote:
>
> > How do you make your RPC do that? It is one of the Acorn pre-defined
> > modes. I always use 1024x768 with 65536 colours. It takes 1536k.
>
> Hey, have I missed something here? My RPC will run in 32K colours and
> 16M colours, but I'm damned if I can get it to run in 64K colours.
Oops! Slip of the calculator. I meant 32k colours of course. 64k colours
would be a little silly.
Talking about screen modes, what would be ideal for piping through to
a TV set? Say you wanted to put a hi-res slideshow on.
--
Phillip Temple
Greg Hennessy
In article <26e54...@tonyh.tcp.co.uk>,
Tony Houghton <to...@tcp.co.uk> writes:
>In article <Dsz92...@systems.DHL.COM>
> ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
>
>> Note I did say 'WWW servers', not the browser. Netscape running over X does
>> not count here. I wasn't the one to make a comparison with the ultra sparc
>> but seeing that someone is trying to make an apples to apples comparison,
>> Any current or future ARM based system is likely to be found wanting as
>> commercial applications or DBMS server platform. NT for Strongarm is
>> just not going to happen.
>
>Point taken, but there is an HTML server for RISC OS. How it compares
>with Netscape I don't know, but all the web pages I've ever seen merely
>consist of a server chucking data at me when asked and occasionally
>running a script. There's no reason the RISC OS version couldn't do
>that.
>
Well here's the list of features of Netscape Enterpise WWW Server
( Courtesy of a cut & paste from www.netscape.com)
Provides advanced capabilities for content creation and management
Including WYSIWYG editing, full text search, and revision control
Extends development platform to include open, server-side applications
First server to support Java and Javascript applications
Increases security and network management capabilities
Including SSL 3.0, client-side certificates, and advanced access control
Support for secure, remote, cross-platform administration, SNMP, and reporting
Employs second generation performance enhancements
Including multi-processor support
A simple www server is very easy to write. However a feature list similiar
to the above on Risc-os is extremely unlikely.
>
>Computers are becoming more data centred than application centred. If
>you're starting with a WWW interface and making it into something that
>depends on proprietary 'standards' you're headed in the wrong
>
>direction. Maybe you had no choice at the time.
>
Well in most organisations that use PC's, Office or something like it is a
'de facto' standard, to say otherwise is being unrealistic :-) . I just use
the WWW to allow the userbase to integrate the facilities they already have
sat on their desk.
>Undoubtedly RISC OS would be better off with PMT, VM and Threading. But
>for some reason every OS that has these features suffers from bloat.
Plase define bloat in this modern age of 1GB hard disk's for ~100 quid
and 16mb of memory for not much more. I work with a broad base of users
that range from someone who can barely switch their machine on to others
who are very switched on :-) and I can say that almost every feature
of the MS-Office applications are being utilised. Including director's
who definitely know their OLE from their elbow. :-)
> VM
>is just a necessity to get them working at all instead of a useful
>enhancement for working with large data, which is now supported on most
>RISC OS applications that need it.
But IIRC on risc-os this facility could be quantified as a 'miserable hack'.
VM is not something to patched on the outside of an OS at some later stage.
Have you ever used a properly configured NT system ? it's a joy to behold (
note no hint of Irony here :-) , Very quick, responsive and performs
well under load. And this on my Dx4/100 with collection of bits at home.
> The lack of modern kernel features
>is mainly a problem to programmers - I am a programmer, yet I still
>forgive it - the user still gets a working environment about as good as
>any for many people's needs.
>
It's these modern features that make life a lot easier for people
like administrators, Ever see BackOffice in action ? Set up an NT system
so Joe User cannot play and therefore make ones life a misery :-).
>>
>> ROFL. You are being slightly naive here. ;-). Check out a Matrox
>> Millenium sometime and then come back to me with the above assertion.
>> Shucks even a 70 quid S3 Trio based board will do the job quite nicely.
>
>My point was that some work requires you to be able to see the result
>of the previous step before you can continue, so the processor being
>available while a redraw is going on isn't much use to you.
Yup it is if you have a PMT system. It spools your print jobs, gets
your mail, Runs your WWW servers etc etc....
> If a Matrox
>Millenium could make redraws instantaneous, then so could a StrongARM
>or good Pentium.
> The difference in performance wouldn't be nearly as
>noticeable as the difference in cost between a dedicated high-
>performance graphics card and relying on a (faster) general-purpose
>processor that you've already got.
>
The difference is quite notable, Note Microsoft's objection to Intel's
attempt to push UMA ( Unified Memory Architecture , A design not unlike
certain systems out of cambridge :-) ) motherboards a few months back.
Then of course there is the whole area of DirectX and 3D, A dedicated
board rendering textured polygons is a lot more sensible than tieing
up the CPU.
>
>The Risc PC was never marketed as a professional graphics workstation,
>so why criticise it for not being as capable as a PC with an expensive
>add-on to make it into one? I don't think a family MPC from a box-
>shifter is likely to have a graphics card that will significantly
>outperform a Risc PC.
>
But again let's make an apples for apples comparison, Given the cost of an
RPC I can have a PC that will quite literally blow it away feature wise an
still have change left over.
>You're obviously no ordinary user don't forget. Most ordinary users
>don't know that they can multi-task at all, so whether it's pre-emptive
>or not doesn't matter to them.
My experience here says otherwise, I have seen secretaries bitching about
the fact that they can run multiple apps under Windows 3.1 with resource
problems. Now they might not know the nitty-gritty with the USER and GDI 64k
heap size but they are clued in well enough to know that they can and do run
multiple windows apps side by side.
> Neither do they need better graphics
>capabilities than what's available with 2MB VRAM and limited bandwidth.
>The Risc PC's more logical interface would suit them better, and if
>more people used it it would generate a market for people who want,
>"What everybody else has got, only better," and push development. It
>would also make the job of people like you (if I've got the right
>impression of what you do) easier. Or perhaps you don't want it to be
>so easy that you're not needed. >:->
>
Again such naivety :-), One has to make use of what one is given, It doesn't
matter whether one has a 1000 mac's, pc's or for that matter though unlikely
:-) 1000 RPC's on users desks out there the adminstration burden does
not go away.
Alex T. Smith
Stuart Halliday sat in the shadows typing :
[ snip ]
> BTW, doesn't a PC video card running a 64,000 colour mode only have a 18-bit
> Palette? I remember a PC programmer complaining about this when he had to
> support 64K colour modes in a DOS program.
There's no reason why not. AFAIK a 65536 (2^16) colour mode is like a
RiscPC 32768 colour mode except that the extra bit is actually added on
to one of the colour components. So it might be 6 bits red, 5 for green
and 5 for blue. A possible complaint is that producing a grey scale isn't
so straight forward.
ObAcorn: Didn't one of the ColourCard or G8 have 65536 capability ?
Alex.
--
Somebody, somewhere is watching Star Trek.
Ian Griffiths
Tony Houghton wrote:
>
> Ian Griffiths <igri...@dev.madge.com> wrote:
>
> > Tony Houghton wrote:
> > >
> > > How are you going to make your already very responsive PC
> > > 4-5x faster for under 300UKP at the end of summer? ;)
> >
> > It's already faster than all my Acorn machines. :-)
>
> Evasion!
OK. I already bought the fast model. I could have bought the
uniprocessor version and then upgraded; as it happens I bought
it with both processor sockets already occupied.
So that doesn't quadruple my speed, but here's my answer to
that:
I can buy a Pentium Pro machine now. I will have to wait until
the end of the summer to get my 4-5x faster RiscPC. If I buy a
StrongARM RiscPC, what can I do THEN to make it 4-5x faster?
The only reason you have that option is that you already bought
the slow machine. I don't currently have a RiscPC, and the one
I intend to get will come with a StrongARM fitted. There won't
be any option for massive speed increase on that because, just
like with my PC, I will already have bought the fast one.
Hell, I could removed both processors from my PC and plug in a
P60, and then say "Behold! Marvel at the impressive
expandability of my machine - I can plug in 2 P166s, and boost
the performance more than fourfold!". Great, thanks, but
frankly I can't be bothered - I'd just like to buy the fast
machine straight away if that's alright. That's just what I'll
be doing when I get my StrongARM-powered RiscPC.
The fact that PC processors have shown strong and steady growth
for years now, whereas ARMs have shown long periods of largely
disappointing change, punctuated by the odd spectacular leap
(The first ARM, ARM3 and StrongARM spring to mind) is not, to my
mind, especially compeeling. The fact is that when the ARM2
appeared, it was, by all available benchmarks, faster than the
opposition (processors in other personal computers), whereas
now, the StrongARM is up there with its competitors, but
certainly not in the lead. The fact that we're seeing a massive
leap seems in this context to indicate that we just had a lot of
catching up to do.
> > Yes 4MB is better than 2. It's not as good as 8. On a 32
> > bit system, it will be slower with 4MB than a PC 4MB
> > graphics card.
>
> Not necessarily, it depends on far too many variables to make
> that generalisation.
Which particular generalisation? More VRAM = better? OK, only
if you actually want to use it.
Speed comparisons? OK, well I'll specify a 4MB 64 bit PC
graphics system which narrows it down, and claim that simple
graphics operations (your basic primitives) will be faster.
This is a narrower claim, but I think it's indisputable. Does
anyone want to dispute it? The wider claim is, admittedly, open
to debate.
> > My point was that for the majority of graphics output, the
> > bus is saturated with existing ARMs. A StrongARM will just
> > be stalled for more clock cycles, it won't actually go any
> > faster, except for stuff which really is slow enough on
> > current machines to be limited by processor speed.
>
> All the evidence proves you wrong.
What evidence would that be? ART's StrongARM test result pages?
Test A - Dhrystones. This has nothing to do with graphics and
is therefore irrelevant.
Tests B and C - JPEG decoding. Not exactly bus limited - JPEG
decode is processor intensive. (Anything that takes on the
order of a second to do what is, when all is said and done,
plotting a sprite to the screen is clearly not being choked by
graphics bandwidth.) The fact that the high colour version runs
faster despite needing to output more bytes shows that the bus
here is not the problem.
Tests D and E - drawing the Artworks Apple. I'll just let my
quoted comments appear:
> > Artworks (once it works) will probably go faster on things
> > like graduated fills.
Evidence proving me wrong is it? :-)
For complex pictures, particularly those involving grad fills
and blends involve quite a lot of processor work not directly
involved with writing the data to the screen.
Conclusion from the data I have: the StrongARM's a lot faster
for Artworks and JPEG work.
What I'm more interested in though is in how fast it is to
redraw and scroll through an Impression document, which in
general involves some simpler output, mostly font-based. (This
is one big reason I have for wanting a graphics accelerator -
these typically hold some of the font cache on the graphics card
and blit directly from there, which can go just stunning fast.)
My guess is that you won't see improvements on anything like the
same scale, if they're noticeable at all.
Of course this is all based on reasoning, rather than
measurement, so I'd be interested to hear real results. The
only ones I've seen so far though are not relevant to what I'm
talking about.
It's entirely possible that I've been overestimating the
efficiency of RiscOS's display code, and that the
processor-video memory bandwidth has never before been a
bottleneck, in which case, all my reasoning will be groundless.
> > get a lot slower when 4MB VRAM machines turn up, unless
> > Acorn do something to make the bus faster.
>
> They said they would. There doesn't seem much point in making
> the bus faster than the RAM, but they are at least going to
> use EDO in future.
That'll help a bit, in that it will further alleviate need for
an L2 cache, but I assume that the video memory will be accessed
at the same speed as before, so that bottleneck remains the
same.
--
Ian Griffiths
Ian Griffiths
Phillip Temple wrote:
>
> Ian Griffiths wrote:
> > I can upgrade my Matrox Millenium to run at 1600x1200 in 16
>
> You can upgrade to 1600x1200 16M?
Yup.
> I make that 1600*1200*4 bytes, which is 7.5Mb of VRAM!
Yup.
> Is this upgrade quite expensive?
Yup.
Comes as a 6MB upgrade, to make a total of 8MB. I'm pretty sure
I can't afford it this quarter, so I don't know what the current
prices are.
--
Ian Griffiths
Ian Griffiths
> > (except perhaps when you're dragging it over complex
> > windows)
>
I don't particularly. I just don't want it to get slower
either.
If you hadn't deleted what I followed on with, it would have
been clear that I was pointing out that doubling the amount of
VRAM, and hence (presumably) doubling the colour depth most
people run at will double the amount of work needing to be done
to move a window around.
At this point you have a choice
1) Put up with windows moving at half the speed they used to
2) Do something to make it faster
In other words, I'm saying that increasing VRAM in isolation
will put more strain on the bus, and result in reduced perceived
performance. A StrongARM won't make much difference because the
bus is the bottleneck in this situation.
If it's still fast enough then I guess that's fine. So long as
big windows still move about as smoothly as the mouse pointer
most of the time then I'm happy. (In other words this is a
theoretical discussion - I don't actually know what window drags
will be like on a 4MB VRAM machine.)
Obviously window drags aren't the only thing - stuff like
filling areas and sprite plotting are in a similar situation.
It's your whole general desktop drawing nuts and bolts I'm
concerned about. Dragging windows around is just the obvious
example. Scrolling is another example, and probably more
relevant.
--
Ian Griffiths
Ian Lynch
Ian Griffiths <igri...@dev.madge.com> wrote:
> If you hadn't deleted what I followed on with, it would have been clear
> that I was pointing out that doubling the amount of VRAM, and hence
> (presumably) doubling the colour depth most people run at will double
> the amount of work needing to be done to move a window around.
This assumes most people work in 24 bit most of the time. I actually
work in 256 colours most of the time because to be honest, spatial
resolution is far more important when writing reports, E-mailing and
reading Usenet which is what I and I suspect many others do most of
the time. I am calling into question whether *most* people need 4 Mb
of VRAM at all. I don't think they do.
> (In other words this is a theoretical discussion - I don't
> actually know what window drags will be like on a 4MB VRAM
machine.)
.
I think we perhaps agree then :-)
--
Ian
Ian Lynch
Ian Griffiths <igri...@dev.madge.com> wrote:
>> Is this upgrade quite expensive?
> Yup.
> Comes as a 6MB upgrade, to make a total of 8MB. I'm pretty sure I
> can't afford it this quarter, so I don't know what the current prices
> are.
If money is no problem why not just by a Silicon Graphics Workstation?
--
Ian
Chris Audley
In article <Dt4tn...@systems.DHL.COM>, Greg Hennessy
<mailto:ghen...@www.lhr-sys.dhl.com> wrote:
>
> In article <26e54...@tonyh.tcp.co.uk>,
> Tony Houghton <to...@tcp.co.uk> writes:
> >In article <Dsz92...@systems.DHL.COM>
> > ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
> >
> >> Note I did say 'WWW servers', not the browser. Netscape running over X does
> >> not count here. I wasn't the one to make a comparison with the ultra sparc
> >> but seeing that someone is trying to make an apples to apples comparison,
> >> Any current or future ARM based system is likely to be found wanting as
> >> commercial applications or DBMS server platform. NT for Strongarm is
> >> just not going to happen.
> >
> >Point taken, but there is an HTML server for RISC OS. How it compares
> >with Netscape I don't know, but all the web pages I've ever seen merely
> >consist of a server chucking data at me when asked and occasionally
> >running a script. There's no reason the RISC OS version couldn't do
> >that.
> >
>
> Well here's the list of features of Netscape Enterpise WWW Server
> ( Courtesy of a cut & paste from www.netscape.com)
>
> Provides advanced capabilities for content creation and management
> Including WYSIWYG editing, full text search, and revision control
Not really anything to do with serving is it - just another app(s) that
give you a better front end to managing the WWW pages, and writing HTML.
There is already a HTML editor available for RISC OS, and what more do you
need than the filer to manage your WWW files :)
> Extends development platform to include open, server-side applications
> First server to support Java and Javascript applications
Haven't really looked at the protocol for serving apps, but I can't imagine
its all that different to serving normal docs. (I have a nasty feeling I'm
going to be proved wrong on that)
> Increases security and network management capabilities
> Including SSL 3.0, client-side certificates, and advanced access control
> Support for secure, remote, cross-platform administration, SNMP, and
> reporting
Quite useful, just a hassle to write, and if you don't need it, it'll just
bloat the app, unless it allows you to 'plug' them in.
Also some of their security stuff is proprietry, so unless someone can hack
it and work out how it works, its unlikely it will appear in any other
server on any platform.
> Employs second generation performance enhancements
Doesn't tell you much, but if its anything like their proxy server, it'll be
using a minimum number of server processes (which have a limited service life
before they die - to stop machines with leaky libraries falling over), and then
it'll start up more processes for transient connections - quite neat, I'm
trying to think of a nice fast way of doing it under RISC OS.
> Including multi-processor support
Mmmmmm, more likely provided by the underlying OS, one feature we probably
won't see for RISC OS :(
> A simple www server is very easy to write. However a feature list similiar
> to the above on Risc-os is extremely unlikely.
Granted, the core of the server with all the 'complicated' sockety stuff
takes very little time to write, its writing all the CGI, Image Map,
authentication and supporting the various headers that takes times - lucky
I've got most of the summer left :)
Note that the Enterprise server costs 995 US dollars, OK so there are cheaper
ones available (cheapest from Netscape is 295 US dollars) but you are still
paying a lot for those extra features - especially when you consider the HTTP
servers for RISC OS are, or are next to, free.
BTW whilst I've been writing software using HTTP, the servers that have
caused me the most hassle by not fully complying to the RFCs have more often
than not been Netscape ones - but then they like setting their own standards
don't they :)
cya
--
Chris Audley, PartII Engineering student, mailto:ch...@santaari.tcp.co.uk
psychotic cyclist & Acorn RiscPC user http://www.tcp.co.uk/~santaari/chris/
Navaho WWW & Proxy Server home page: http://www.tcp.co.uk/~santaari/navaho/
Ian Lynch
Pe...@hstein.demon.co.uk (Peter Smith) wrote:
> Similarly 24bit colour takes up 1876k. This is because there are 8 bits
> each of red, green and blue, and an 8 bit alpha channel. Well handy.
> (Alledgedly :-)
Certainly is for TV and video.
--
Ian
Tony Houghton
In article <Dt4tn...@systems.DHL.COM>
ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
> Well here's the list of features of Netscape Enterpise WWW Server
> ( Courtesy of a cut & paste from www.netscape.com)
>
> Provides advanced capabilities for content creation and management
> Including WYSIWYG editing, full text search, and revision control
> Extends development platform to include open, server-side applications
> First server to support Java and Javascript applications
> Increases security and network management capabilities
> Including SSL 3.0, client-side certificates, and advanced access
> control
> Support for secure, remote, cross-platform administration, SNMP, and
> reporting
> Employs second generation performance enhancements
> Including multi-processor support
>
>
> A simple www server is very easy to write. However a feature list similiar
> to the above on Risc-os is extremely unlikely.
The fact that RISC OS doesn't have such a server isn't a disaster for
the platform. There's nothing to stop you connecting RISC OS clients to
NT servers. I do agree that that would be far more sensible than using
RISC OS to run a server. A server can benefit from features of Windows
NT, but not from features of RISC OS, namely its better GUI.
> Well in most organisations that use PC's, Office or something like it is a
> 'de facto' standard, to say otherwise is being unrealistic :-) . I just use
> the WWW to allow the userbase to integrate the facilities they already have
> sat on their desk.
It's a shame they can get away with calling something a standard when
its file formats and source code etc are so jealously guarded.
> >Undoubtedly RISC OS would be better off with PMT, VM and Threading. But
> >for some reason every OS that has these features suffers from bloat.
>
> Plase define bloat in this modern age of 1GB hard disk's for ~100 quid
> and 16mb of memory for not much more. I work with a broad base of users
> that range from someone who can barely switch their machine on to others
> who are very switched on :-) and I can say that almost every feature
> of the MS-Office applications are being utilised. Including director's
> who definitely know their OLE from their elbow. :-)
I'd define bloat as something that has :
> secretaries bitching about
> the fact that they can run multiple apps under Windows 3.1 with resource
> problems
There isn't much advantage to not needing much disc space nowadays
(except they last longer through not being thrashed so much, but that's
irrelevant to PCs), but you can get 4MB SIMMs for about 30UKP now.
That's better than having to spend >100UKP for the same functionality
isn't it?
> But IIRC on risc-os this facility could be quantified as a 'miserable hack'.
So what? It does the job, there's nothing miserable about it for the
user who doesn't care what's "under the bonnet".
The whole of Windows could be quantified as a 'miserable hack' on top
of DOS.
> >My point was that some work requires you to be able to see the result
> >of the previous step before you can continue, so the processor being
> >available while a redraw is going on isn't much use to you.
>
> Yup it is if you have a PMT system. It spools your print jobs, gets
> your mail, Runs your WWW servers etc etc....
I don't think there's an overpowering majority of people who do all
those things at once just because they can. Most people would rather
have all the power of the machine concentrated on their demanding
graphics job. Are you going to tell me that NT doesn't slow down the
more tasks it runs, and also claim that its (multi-processor) thread
scheduling makes best use of all configurations?
> The difference is quite notable, Note Microsoft's objection to Intel's
> attempt to push UMA ( Unified Memory Architecture , A design not unlike
> certain systems out of cambridge :-) ) motherboards a few months back.
No, I didn't note it. What is UMA? I think you're being rather naive if
you believe that Microsoft wanted to avoid rewriting their kernel for a
new architecture solely because the old one was better.
> Then of course there is the whole area of DirectX and 3D, A dedicated
> board rendering textured polygons is a lot more sensible than tieing
> up the CPU.
Must be very useful for MS Office.
> Again such naivety :-), One has to make use of what one is given, It doesn't
> matter whether one has a 1000 mac's, pc's or for that matter though unlikely
> :-) 1000 RPC's on users desks out there the adminstration burden does
> not go away.
Again, such naivety :-) Some platforms do create more problems than
others. Maybe NT is relatively trouble-free when properly set up, but
that's because it was written for power users who need something bullet
proof and are likely to be good at setting it up properly. How many
secretaries outside your company use it for word-processing etc? The
vast majority use Windows 3.1 or 95, which were written for people who
will just use what they're told to and won't be disappointed when the
computer lives up to their expectations of having a mind of its own.
Tony Houghton
In article <31C553...@dev.madge.com>
Ian Griffiths <igri...@dev.madge.com> wrote:
> Tony Houghton wrote:
>
> Hell, I could removed both processors from my PC and plug in a
> P60, and then say "Behold! Marvel at the impressive
> expandability of my machine - I can plug in 2 P166s, and boost
> the performance more than fourfold!". Great, thanks, but
> frankly I can't be bothered - I'd just like to buy the fast
> machine straight away if that's alright. That's just what I'll
> be doing when I get my StrongARM-powered RiscPC.
Oh, so you are going to buy one? Sorry, I thought you were rubbishing
them.
> > > Yes 4MB is better than 2. It's not as good as 8. On a 32
> > > bit system, it will be slower with 4MB than a PC 4MB
> > > graphics card.
> >
> > Not necessarily, it depends on far too many variables to make
> > that generalisation.
>
> Which particular generalisation? More VRAM = better? OK, only
> if you actually want to use it.
>
> Speed comparisons? OK, well I'll specify a 4MB 64 bit PC
> graphics system which narrows it down, and claim that simple
> graphics operations (your basic primitives) will be faster.
> This is a narrower claim, but I think it's indisputable. Does
> anyone want to dispute it? The wider claim is, admittedly, open
> to debate.
Wouldn't simple primitives be the sort of things that would fit nicely
in the StrongARM's caches? Perhaps you mean large blits and things, in
which case I'd agree with you. But if you meant line drawing primitives
I'd disagree.
> Conclusion from the data I have: the StrongARM's a lot faster
> for Artworks and JPEG work.
>
> What I'm more interested in though is in how fast it is to
> redraw and scroll through an Impression document, which in
> general involves some simpler output, mostly font-based. (This
> is one big reason I have for wanting a graphics accelerator -
> these typically hold some of the font cache on the graphics card
> and blit directly from there, which can go just stunning fast.)
> My guess is that you won't see improvements on anything like the
> same scale, if they're noticeable at all.
Someone said that he tried changing the base font size of a large
Impression document at one of the shows and the update speed was
amazing.
> Of course this is all based on reasoning, rather than
> measurement, so I'd be interested to hear real results. The
> only ones I've seen so far though are not relevant to what I'm
> talking about.
That seems true, but if the only tests that had been done were thing
that were expected to be fast anyway, there wouldn't be so much
delighted surprise at the better than expected performance. Or perhaps
it's just that tasks that require good bus bandwidth more than anything
else aren't really used on this platform, which would make it true of
most mid-range PC users as well.
> It's entirely possible that I've been overestimating the
> efficiency of RiscOS's display code, and that the
> processor-video memory bandwidth has never before been a
> bottleneck, in which case, all my reasoning will be groundless.
The fact that it has to cope with things like mode-independence and
rectangles that don't have their boundaries word-aligned would add a
bit to the amount of code having to be executed.
An alternative theory is that even with the bottleneck, things that
rely on the bandwidth and not on CPU speed can already be done in the
blink of an eye anyway, so having a bottleneck doesn't really matter.
The stuff that we usually have to wait for is calculation of what to
plot where, and that's what the StrongARM is good at.
Iain Anderson
Tony Houghton (to...@tcp.co.uk) wrote:
: In article <31C012...@dev.madge.com>
: Ian Griffiths <igri...@dev.madge.com> wrote:
: How are you going to make your already very responsive PC 4-5x faster
: for under 300UKP at the end of summer? ;)
Tee hee. How about "buy 32MB of DRAM". This could well make the machine 5x
faster under certain conditions :)
: > Furthermore, the lack of a graphics accelerator will
: > start to be a problem when more VRAM is available, because the
: > processor will be having to work twice as hard, and since for
: > graphics it's mostly bus limited, the StrongARM won't make
: > things any better.
: It will in fact. My money's on the StrongARM being faster than a Matrox
: Millenium. I don't think that just because it's a dedicated graphics
: card it means it can magically handle 4-8MB VRAM just as fast as 1-2MB.
Can anyone think of a good way of making this comparison ? I would guess that
the Matrox would outperform the SA, but it would be interesting to see how
they compare - and it would certainly be interesting if Tony is right.
____
Iain
Iain Anderson
Ian Griffiths (igri...@dev.madge.com) wrote:
: Tower Electronics Ltd wrote:
: >
: > Ian Griffiths <mailto:igri...@dev.madge.com> wrote:
: > >
: > > Why exactly? There are already 3 high performance RISC
: > > architectures around which will run NT.
: >
: > AND HAVE YOU LOOKED AT THE PRICE TAG - go on then purchase
: Well the Sun won't run NT of course, and neither will the HP. I
: didn't think that PowerPC machines were that expensive.
: PowerMACs certainly aren't.
: > and see how much change you get out of 60,000 ukp!
: For a PowerMAC? Quite a lot actually.
: And the next generation PowerMACs will be able to run NT.
: I don't have any appropriate price lists to hand. Can you tell
: me how much a PPC-based PC costs? I was guessing at sub-2000
: UKP. I'm certain it's not 65k.
An example that we've just got :
RS/6000 43P
PowerPC 604 at 133MHz
128MB DRAM
2MB VRAM
20" monitor
Ethernet
SCSI-2, with internally CD-ROM and 2 of 2GB hard drives
AIX (blech), Solaris and Windows NT.
Lease at 2k UKP a year or buy for about 15k UKP. Bit of an easy decision
there. And yes, even with IBM's margins it's a sight less than 60 thou.
____
Iain
Greg Hennessy
In article <b182d...@tonyh.tcp.co.uk>,
Tony Houghton <to...@tcp.co.uk> writes:
>> to the above on Risc-os is extremely unlikely.
>
>The fact that RISC OS doesn't have such a server isn't a disaster for
>the platform. There's nothing to stop you connecting RISC OS clients to
>NT servers. I do agree that that would be far more sensible than using
>RISC OS to run a server. A server can benefit from features of Windows
>NT, but not from features of RISC OS, namely its better GUI.
>
Hmmm.. I don't know, NT 4.0 is pretty slick :-). You might even
be mildly impressed :-)
>
>It's a shame they can get away with calling something a standard when
>its file formats and source code etc are so jealously guarded.
>
Well C'est La Vie, It helps pay the rent for a lot of us so don't knock it
:-)
>I'd define bloat as something that has :
>
>> secretaries bitching about
>> the fact that they can run multiple apps under Windows 3.1 with resource
>> problems
>
Well that really has nothing got to do with bloat, The problem exists due to
the architectural limitations of windows 3.x. Which to give a small
bit of kudos to '95 has largely disappeared.
>There isn't much advantage to not needing much disc space nowadays
>(except they last longer through not being thrashed so much, but that's
>irrelevant to PCs), but you can get 4MB SIMMs for about 30UKP now.
>That's better than having to spend >100UKP for the same functionality
>isn't it?
>
Well you should see the cost of a site license for something like MS Office,
It's a lot less than you think per place setting. The same with items like
Netscape Navigator.
>> But IIRC on risc-os this facility could be quantified as a 'miserable hack'.
>
>So what? It does the job, there's nothing miserable about it for the
>user who doesn't care what's "under the bonnet".
>
But you being the application developer :-) would agree that applications
should not have to be specially engineered to use something that should
be there as standard.
>The whole of Windows could be quantified as a 'miserable hack' on top
>of DOS.
>
:-) I won't disagree there :-)
>I don't think there's an overpowering majority of people who do all
>those things at once just because they can. Most people would rather
>have all the power of the machine concentrated on their demanding
>graphics job.
IMHO it's a question of perception, The 'overpowering majority of people'
generally do not do 'demanding graphics jobs'. It's usually the more
mundane stuff like Word, XL etc.
Are you going to tell me that NT doesn't slow down the
>more tasks it runs, and also claim that its (multi-processor) thread
>scheduling makes best use of all configurations?
>
Nah the next thing I will assert is that NT is so fast it does an infinite
loop in 5 seconds :-) .But seriously it depends on the task, You will find
that the majority of software for most of the runtime is I/O not compute
bound. NT performs quite well under load. But as you say it's a question of
horses for courses.
>> The difference is quite notable, Note Microsoft's objection to Intel's
>> attempt to push UMA ( Unified Memory Architecture , A design not unlike
>> certain systems out of cambridge :-) ) motherboards a few months back.
>
>No, I didn't note it. What is UMA? I think you're being rather naive if
>you believe that Microsoft wanted to avoid rewriting their kernel for a
>new architecture solely because the old one was better.
>
IIRC Intel are pushing UMA as a cheaper way to make motherboards. UMA
boards have the System and the Video controller sharing the same memory.
Needless to say as resolution an colour depth increased system performance
goes out the window ( figuratively speaking :-) )
>> Then of course there is the whole area of DirectX and 3D, A dedicated
>> board rendering textured polygons is a lot more sensible than tieing
>> up the CPU.
>
>Must be very useful for MS Office.
>
Well I am not the one wanting to tie up the CPU to render graphics :-)
>
>Again, such naivety :-) Some platforms do create more problems than
>others. Maybe NT is relatively trouble-free when properly set up, but
>that's because it was written for power users who need something bullet
>proof and are likely to be good at setting it up properly.
Well you may have to get used to it. I hear a lot of noises from companies
that are going to skip the transition from win 3.x to '95 and go direct to
NT 4.0.
>How many
>secretaries outside your company use it for word-processing etc? The
>vast majority use Windows 3.1 or 95, which were written for people who
>will just use what they're told to and won't be disappointed when the
>computer lives up to their expectations of having a mind of its own.
>
Well given that assertion would you not agree that the future for
a platform like the RPC is rather bleak.
Toby Haynes
In article <31C553...@dev.madge.com>,
Ian Griffiths <igri...@dev.madge.com> wrote:
>> > My point was that for the majority of graphics output, the
>> > bus is saturated with existing ARMs. A StrongARM will just
>> > be stalled for more clock cycles, it won't actually go any
>> > faster, except for stuff which really is slow enough on
>> > current machines to be limited by processor speed.
>> All the evidence proves you wrong.
>What evidence would that be? ART's StrongARM test result pages?
>Test A - Dhrystones. This has nothing to do with graphics and
>is therefore irrelevant.
Indeed - dhrystones is a good test for the CPU in *isolation* to it's
surroundings - it's useless at comparing complete systems. It is still
possible to inflate a CPU's dhrystones results with specific compiler
optimisations so it's not unbiased from a compiler point of view
either. I suspect that several compilers are dhrystone-enhanced to
make them more attractive in what is after all a fairly widely known
benchmark. I don't believe that Acorn Cv5 has any specific dhrystone
compiler options though ... ;-)
>Tests B and C - JPEG decoding. Not exactly bus limited - JPEG
>decode is processor intensive. (Anything that takes on the
>order of a second to do what is, when all is said and done,
>plotting a sprite to the screen is clearly not being choked by
>graphics bandwidth.) The fact that the high colour version runs
>faster despite needing to output more bytes shows that the bus
>here is not the problem.
>Tests D and E - drawing the Artworks Apple. I'll just let my
>quoted comments appear:
>> > Artworks (once it works) will probably go faster on things
>> > like graduated fills.
>Evidence proving me wrong is it? :-)
Absolutely not! ;-)
[snip]
>Conclusion from the data I have: the StrongARM's a lot faster
>for Artworks and JPEG work.
>What I'm more interested in though is in how fast it is to
>redraw and scroll through an Impression document, which in
>general involves some simpler output, mostly font-based. (This
>is one big reason I have for wanting a graphics accelerator -
>these typically hold some of the font cache on the graphics card
>and blit directly from there, which can go just stunning fast.)
>My guess is that you won't see improvements on anything like the
>same scale, if they're noticeable at all.
Having stood and played around with the StrongARM machine at the ARM
'96 show in Cambridge, I would say that the improvement in speed in
window redrawing is not only noticeable, but dramatic. Having watched
Chris Cox drag windows all over the place with barely a noticeable
judder (it only really became noticeable in the fraction of a second
it redraws Draw windows) I was impressed. I also was impressed with
the Tumble Demo, which if my coding is anything to go by, should be
spending a large part of it's time reading and writing to
memory. Similarly, people who got to play with Style on the machine at
Wakefield reported that it was silky smooth in scrolling a 70 page
document - including pictures and text. This certainly is encouraging
- and maybe points again to Risc OS and apps not being as hard on
memory accesses as was feared. Or maybe that DSP Multiply unit is
doing stormingly good work.
Of course, nobody has had their hands on the machine (outside Acorn of
course) for long enough to discover where the bottlenecks are now in
the system. Whether the improvements in scrolling speed are up to the
4-5 times for some of the other benchmarks is also not clear. I also
think that as the amount of VRAM steps up, a graphics accelerator/
caching system will become increasingly necessary to prevent things
crawling in memory intensive modes. In 8Mb modes (ie 1600x1200 @
32bpp) you are throwing a huge amount of memory around (about 500MB/s)
- both memory bandwidth and processor bandwidth need to be superb to
cope with this at these sorts of resolutions, and our Risc PC's won't
cope at the moment. Since this is not the sort of thing that everyone
is going to want/need, it shouldn't comprise a fundamental part of the
motherboard - it's far better that it is in a high-speed podule with
fast access into the rest of the system, and need only be bought by
those who have to have it.
Cheers,
Toby
--
Toby Haynes | "I COULD MURDER A CURRY" - Death
Somewhere in Cambridge | Mort by Terry Pratchett
Tony Houghton
In article <Dt6qz...@systems.DHL.COM>
ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
> In article <b182d...@tonyh.tcp.co.uk>,
> Tony Houghton <to...@tcp.co.uk> writes:
> >> to the above on Risc-os is extremely unlikely.
> >
> >The fact that RISC OS doesn't have such a server isn't a disaster for
> >the platform. There's nothing to stop you connecting RISC OS clients to
> >NT servers. I do agree that that would be far more sensible than using
> >RISC OS to run a server. A server can benefit from features of Windows
> >NT, but not from features of RISC OS, namely its better GUI.
> >
>
> Hmmm.. I don't know, NT 4.0 is pretty slick :-). You might even
> be mildly impressed :-)
Is that the new version whose interface is more like Windows 95? If it
is, then it must be quite nice.
> >There isn't much advantage to not needing much disc space nowadays
> >(except they last longer through not being thrashed so much, but that's
> >irrelevant to PCs), but you can get 4MB SIMMs for about 30UKP now.
> >That's better than having to spend >100UKP for the same functionality
> >isn't it?
> >
>
> Well you should see the cost of a site license for something like MS Office,
> It's a lot less than you think per place setting. The same with items like
> Netscape Navigator.
What's that got to do with their RAM requirements?
> But you being the application developer :-) would agree that applications
> should not have to be specially engineered to use something that should
> be there as standard.
I don't have to specifically engineer my applications for VM. All I need
is a good memory management library using Dynamic Areas while being
backwards compatible with older systems. I've just been writing one. You
can even patch the Standard C Library to use Dynamic Areas.
> IMHO it's a question of perception, The 'overpowering majority of people'
> generally do not do 'demanding graphics jobs'. It's usually the more
> mundane stuff like Word, XL etc.
They don't need a powerful graphics accelerator either.
> IIRC Intel are pushing UMA as a cheaper way to make motherboards. UMA
> boards have the System and the Video controller sharing the same memory.
> Needless to say as resolution an colour depth increased system performance
> goes out the window ( figuratively speaking :-) )
You mean with no VRAM? That is a bit daft for PCs when you can get dirt
cheap graphics cards. Unfortunately Acorn didn't have that option when
cost-reducing the A7000, but its capabilities are adequate for schools.
> Well you may have to get used to it. I hear a lot of noises from companies
> that are going to skip the transition from win 3.x to '95 and go direct to
> NT 4.0.
Isn't that because you can't easily upgrade from Win 95 to NT? Something
to do with the former's long filenames being a hack.
> Well given that assertion would you not agree that the future for
> a platform like the RPC is rather bleak.
It was looking pretty bleak until the likes of Oracle got interested.
Acorn's attitude has completely changed and they should be able to
exploit their technology outside the dwindling education market now. The
situation is made to look worse by the increased popularity of PCs
rather than decreased popularity of Acorns.
Greg Hennessy
In article <362e5...@tonyh.tcp.co.uk>,
Tony Houghton <to...@tcp.co.uk> writes:
>> Hmmm.. I don't know, NT 4.0 is pretty slick :-). You might even
>> be mildly impressed :-)
>
>Is that the new version whose interface is more like Windows 95? If it
>is, then it must be quite nice.
>
Yup, thats the one. It's quite impressive.
>> Well you should see the cost of a site license for something like MS Office,
>> It's a lot less than you think per place setting. The same with items like
>> Netscape Navigator.
>
>What's that got to do with their RAM requirements?
>
Well in most commercial environments you will find that Hardware is leased
due to tax reasons therefore the cost analysis of putting decent spec'd
hardware on desks to run this software is quite different when compared to
purchasing a system outright. As regards the software you then run,
regardless of RAM requirements if it's not a 'brand' name with all that
implies like support contracts, SLA's etc then the bean counters are not
going to authorise its purchase.
>> But you being the application developer :-) would agree that applications
>> should not have to be specially engineered to use something that should
>> be there as standard.
>
>I don't have to specifically engineer my applications for VM. All I need
>is a good memory management library using Dynamic Areas while being
>backwards compatible with older systems. I've just been writing one. You
>can even patch the Standard C Library to use Dynamic Areas.
>
Ahh but you have had to specifically alter/create code to make use of this,
Over the years ( Christ I feel old :-) ) I have knocked up a fair bit
of code an systems like VMS, Unix of varying sorts, NT, & OS2 and I
haven't had to write a memory management library yet :-)
>> IMHO it's a question of perception, The 'overpowering majority of people'
>> generally do not do 'demanding graphics jobs'. It's usually the more
>> mundane stuff like Word, XL etc.
>
>They don't need a powerful graphics accelerator either.
Well you will find that stock systems from vendors like Gateway, HP, Compaq
and Dell are shipping as standard with products like the Millenium and the
#9 Motion FX 771. Kinda up's the ante for non pc systems trying to compete.
>
>
>You mean with no VRAM? That is a bit daft for PCs when you can get dirt
>cheap graphics cards. Unfortunately Acorn didn't have that option when
>cost-reducing the A7000, but its capabilities are adequate for schools.
>
Yup, thats the story, Hence Microsoft reluctance, But anyone with half
a brain are leaving systems based on this architecture on the shelf and
going for the usual PCI based graphics accelerator.
>
>Isn't that because you can't easily upgrade from Win 95 to NT? Something
>to do with the former's long filenames being a hack.
>
IIRC the long filenames are not the problem, The hassle is with the
'95 registry. It's completly incompatible with the NT registry hence
there is no means of migrating the settings for any software installed
under '95 to NT. A big problem if you have a couple of thousand
MS-Office users out there.
>
>It was looking pretty bleak until the likes of Oracle got interested.
>Acorn's attitude has completely changed and they should be able to
>exploit their technology outside the dwindling education market now. The
>situation is made to look worse by the increased popularity of PCs
>rather than decreased popularity of Acorns.
>
Hmm... I don't know. It looks to me that the only winners here are
likely to be those have speculated in Acorn shares over the past 12
months.
Tony Houghton
In article <Dt8I6...@systems.DHL.COM>
ghen...@www.lhr-sys.dhl.com (Greg Hennessy) wrote:
> Yup, thats the story, Hence Microsoft reluctance, But anyone with half
> a brain are leaving systems based on this architecture on the shelf and
> going for the usual PCI based graphics accelerator.
The only reason separate graphics accelerators are cost effective for
bottom-end PCs is because of the economies of huge scale. Separate
graphics cards were made necessary because of deficiencies in early PCs'
hardware. I agree that VRAM is essential in this day and age, but it
would make sense to ship low-end PCs with a system similar to the Risc
PC's. It would be cheaper, and you'd still have the option of upgrading.
It would be possible to add a graphics accelerator to a Risc PC, but
because its native capabilities aren't entirely hopeless there isn't the
demand.
> Hmm... I don't know. It looks to me that the only winners here are
> likely to be those have speculated in Acorn shares over the past 12
> months.
All that seems to matter in the software market is size of company.
Seeing as Oracle are second only to Microsoft they should do OK for us.
Clifton College
"Alex T. Smith" <A.T....@ncl.ac.uk> wrote:
>There's no reason why not. AFAIK a 65536 (2^16) colour mode is like a
>RiscPC 32768 colour mode except that the extra bit is actually added on
>to one of the colour components. So it might be 6 bits red, 5 for green
>and 5 for blue. A possible complaint is that producing a grey scale isn't
>so straight forward.
But look at the 8-bit modes on Archimedes machines. Two bits red, two
bits green, two bits blue and four levels of "tint" - a small amount
added to each of the three.
So you get 8 bit colour with 16 greys. There's no reason why the 16th
bit couldn't be used similarly to lighten or darken colours.
AFAIK this was also used by the Spectrum! Each colour was four bits,
including the option to make it "bright". So it's a pretty old idea.
Ed Avis