ARM and DEC developing StrongARM
Manar Hussain
--
Embargoed Until Monday February 6, 1995 UK62
<hmm - didn't stop the Sunday Times... - ed>
Digital and ARM Announce Agreement to Develop
High-Speed, Low-Power, ARM-Compliant Processors
... World's Performance Leader Teams with Power-Efficiency Leader
Maynard MA and Cambridge UK -- February 6, 1995 --
Digital Equipment Corporation and Advanced RISC Machines Ltd (ARM)
today announced the licensing of the ARM RISC architecture to Digital
Semiconductor for the development of high-performance, low power
microprocessors. The StrongARM family of 32-bit RISC products to be
developed under the agreement is intended to complement and broaden the
existing ARM product line for performance critical applications such as:
- next-generation personal digital assistants (PDAs) with improved user
interfaces and communications,
- interactive TV and set-top products,
- video games and multimedia 'edutainment' systems with realistic
imaging, motion and sound, and
- digital imaging, including low cost digital image capture and
photo-quality scanning and printing.
Combining Digital's proven leadership in high performance
microprocessor design and manufacture with ARM's expertise in low
power design, will result in processors that set a new standard
for high performance while meeting the low power, space and cost
requirements of products such as handheld devices. The StrongARM
family will offer full software compatibility with the ARM6, 7 and
8 chip families, which will help accelerate market acceptance of
the new products.
Strengthens Merchant Vendor and Performance Commitment
For Digital Semiconductor, this is a strategic agreement that both
reinforces our merchant vendor role and demonstrates performance
leadership, said Ed Caldwell, vice president and general manager
of Digital Semiconductor. "Today, our Alpha products provide
unmatched performance for desktop and server applications. The
StrongARM product line will complement this strategy with its focus
on enhancing performance for mass-market applications in which very
low power dissipation is critical."
"This agreement with ARM also gives us early entry into rapidly
growing, high volume markets" Caldwell added. Industry analysts
estimate that the market for 32-bit RISC embedded consumer
applications will grow 75 percent year over year to more than
10.5 million in 1998 (InStat October 1994).
Performance is Key
According to Robin Saxby, Managing Director & CEO of ARM: "Having
Digital Semiconductor jointly design and build new processors
compliant with the ARM architecture will add momentum to ARM's
acceptance as the volume RISC standard for 32-bit applications.
ARM processors already have the best ratios of performance to
power consumption and cost (MIPS/Watt and MIPS/$).
The agreement with Digital will maintain our lead in these areas
while allowing us to pursue applications demanding very high
absolute performance."
Shane Robison, vice president and general manager of Apple
Computer, Inc's Personal Interactive Electronics Division, said
Apple was an early adopter of ARM microprocessor technology and
had incorporated the ARM 610 processor into its market-leading
Newton MessagePad PDA. "Apple's Newton engineering team has
been working closely with Digital Semiconductor and ARM in
defining the first StrongARM microprocessor. This design looks
to significantly boost compute performance while retaining the
low power characteristic of ARM microprocessors, both of which
are critical in designing high performance PDAs."
Sam Wauchope, the Acorn Computer Group Managing Director
commented that "Acorn was the first personal computer
manufacturer to move to RISC in 1987 after it developed the
original ARM processor. The announcement of StrongARM ensures
that our powerful, low-cost multimedia workstations for education,
the home and professional use will continue to have the leading
performance to deliver to our customers the innovative solutions
they are demanding."
The Chief Executive of interactive TV set-top box designer Online
Media, Malcolm Bird, said "The ARM processor was the obvious
choice for our digital set-top products. The new StrongARM devices
will offer exciting new possibilities for performance and in
hardware/software trade-offs. We already have set-top products in
production and will be able to rapidly benefit from the new devices."
"This relationship looks to be a perfect strategic fit," said Jerry
Banks, Director/Principal Analyst, Dataquest. "ARM gains access to
high performance microprocessor design and process technology, while
Digital gains ARM's expertise in low power design, as well as access
to high volume markets with significant potential. The resulting
products could have a far-reaching effect on many emerging consumer
applications like PDAs, interactive TV, and games."
First product in development today
The first product in the StrongARM family is currently under
development at Digital Semiconductor's Palo Alto, California, and
Austin, Texas, research centres and ARM's Cambridge, UK headquarters.
Digital expects the device to be among the first products manufactured
at its new FAB 6 state-of-the-art chip fabrication facility in Hudson,
Massachusetts. The products developed under the agreement will be
sold through Digital Semiconductor's sales channels. In addition,
processors and processor cores developed under this agreement will be
available for licensing to other semiconductor partners. Saxby added
"This is consistent with our strategy of making the ARM architecture
an open standard for performance oriented, power-efficient and cost-
effective applications."
-----
comments:
a) Well all you guys shouting about why ARM don't push the performance
envelope a bit - shut up ;)
b) Does anyone have a rough idea of when the new DEC FAB 6 in Hudson is
due to open - that gives a strong clue as to when we will first see
some StrongARMs...
c) I guess this explains where all those rumours about Newtons switching
to PowerPC chips came from...
d) Wheehayy.... ;)
--
Manar Hussain RiscBSD Co-ordinator
ma...@iVision.demon.co.uk unix for the Risc PC
Graduate Horizons - The UK Careers Information Resource - is looking for
contributions (eg career profiles) - http://ww.gold.net/arcadia/horizons
Manar Hussain
> ma...@iVision.demon.co.uk unix for the Risc PC
In case you're wondering, there should be a clarification of the Arc
unix position rsn (I mean rrsn)....
> Graduate Horizons - The UK Careers Information Resource - is looking for
> contributions (eg career profiles) - http://ww.gold.net/arcadia/horizons
Sorry to have to correct this, it's - http://www.gold.net/arcadia/horizons/
Chow,
Manar
Beej
: In article <792031...@ivision.demon.co.uk> Ma...@ivision.demon.co.uk (Manar Hussain) writes:
: > Digital and ARM Announce Agreement to Develop
: > High-Speed, Low-Power, ARM-Compliant Processors
: Yahoo! Whoop! Whoop! Whoop! <somersault> Yeehaaar! Ekee Ekee Ekee Zoooooooom!
: I do believe that's the best news I've heard all year. Especially the full
: software compatibility bit.
: Kevin
: =====
Digital are a little down in the doldrums still though vis a vis money.
Bob
: > Digital and ARM Announce Agreement to Develop
: > High-Speed, Low-Power, ARM-Compliant Processors
Shouldn't this be in comp.sys.arm?
Neil?
Thilo Manske
>In article <792031...@ivision.demon.co.uk> Ma...@ivision.demon.co.uk (Manar Hussain) writes:
>> Digital and ARM Announce Agreement to Develop
>> High-Speed, Low-Power, ARM-Compliant Processors
>Yahoo! Whoop! Whoop! Whoop! <somersault> Yeehaaar! Ekee Ekee Ekee Zoooooooom!
>I do believe that's the best news I've heard all year. Especially the full
>software compatibility bit.
>Kevin
>=====
Oh , YES!!!
We will get a machine that can compete with a 90MHz P5 in five years - in
benchmarks, but Intel will have finished their work (and errata) with a
200MHz P7 by this time - driving >>75 Watts ;-) at 2 Volts, >20M transistors,
this will be enough speed to run Windows'97's 16Bit code as fast as RiscOS-
software on a A5k these days.
Oh, ARM Ltd./Acorn come on! A simple 55MHz ARM 700 + FPA card will be
sufficient for most of the users, and I don't think this is impossible
at the moment since there were such cards shown on different shows!
I think, the only reason for this deal is, that DEC has noticed, that their
Alpha was a big fault (as the PowerPC, these CPUs have more than 200
instructions each and call themselves RISC! The only advantages are bigger
register sets (32 AFAIK), superscalar-architecure (the PowerPC can execute
max. 3 instructions in one cycle), floating-point-instructions and higher
clock-rates, but this costs them a lot of transistors and a lot of Watts - they
are as hot as P90's! )
But let's wait, what will come... We Acorn-users allways have to wait :-(
--
------------------------------------------------------------------------------
Thilo....@Informatik.Uni-Oldenburg.DE
RiscPC600 ARM610 @33MHz 12+1MB RAM @16MHz 420MB IDE-HD RO3.50D
BBC A3000 ARM 2 @ 8MHz 4MB RAM @ 8MHz 250MB IDE-HD RO3.19 GU
I know which side of the soldering iron is the hot one...
*OUCH*
Sorri four may bed English.
Jon Ribbens
> Shouldn't this be in comp.sys.arm?
Err, I think that the idea is that hopefully it will have implications
for *Acorn* computers. (400MHz RISC PC anybody? Now wouldn't that be
nice ;-) )
Jon
--
// Jon Ribbens // es...@csv.warwick.ac.uk //
// http://www.csv.warwick.ac.uk/~esveb/ //
Ben Laughton
yet :-)
I heard a while ago that if current ARM's were thrown down the really
modern processor manufacturing lines (does this imply a lower micron
thingy?) then we would see an instant performance imporovement. (Is
this because we'd be able to clock them faster - my knowledge in this
area isn't great!).
Ale Terlevich
>Figured this would be of interest..
>
>--
>Embargoed Until Monday February 6, 1995 UK62
>
><hmm - didn't stop the Sunday Times... - ed>
>
> Digital and ARM Announce Agreement to Develop
> High-Speed, Low-Power, ARM-Compliant Processors
>
>
>
>Performance is Key
>
>Shane Robison, vice president and general manager of Apple
>Computer, Inc's Personal Interactive Electronics Division, said
>Apple was an early adopter of ARM microprocessor technology and
>had incorporated the ARM 610 processor into its market-leading
>Newton MessagePad PDA. "Apple's Newton engineering team has
>been working closely with Digital Semiconductor and ARM in
>defining the first StrongARM microprocessor. This design looks
>to significantly boost compute performance while retaining the
>low power characteristic of ARM microprocessors, both of which
>are critical in designing high performance PDAs."
>c) I guess this explains where all those rumours about Newtons switching
>to PowerPC chips came from...
>
Huh? I don't get that one.
This seems to strongly imply that Apple Newts are going to keep using ARMs,
albeit StrongARMs ( Wow! What a comic name :-) )
Personally I like the bit above where it sais "This design looks to
significantly boost compute performance"
:-)
It would be nice if Acorn (you listening?) could put some info about it
in this clan pack that's meant to me coming real soon now...
Ale.
M J Ebourne
K.J. Bracey <kjb...@phx.cam.ac.uk> wrote:
>In article <792031...@ivision.demon.co.uk> Ma...@ivision.demon.co.uk (Manar Hussain) writes:
>
>> High-Speed, Low-Power, ARM-Compliant Processors
>
>
>I do believe that's the best news I've heard all year. Especially the full
>software compatibility bit.
Not that we've exactly had much year yet...
Does sound good though. I want one urgently. No, make it two. :-)
--
Martin Ebourne "Pray that there's intelligent
Electronics, life somehere out in space,
Southampton University because there's bugger all
Email: m...@soton.ac.uk down here on earth." - Python
Ale Terlevich
In article 000A...@phx.cam.ac.uk, kjb...@phx.cam.ac.uk (K.J. Bracey) writes:
>In article <792031...@ivision.demon.co.uk> Ma...@ivision.demon.co.uk (Manar Hussain) writes:
>
>> Digital and ARM Announce Agreement to Develop
>> High-Speed, Low-Power, ARM-Compliant Processors
>
>Yahoo! Whoop! Whoop! Whoop! <somersault> Yeehaaar! Ekee Ekee Ekee Zoooooooom!
>
>I do believe that's the best news I've heard all year. Especially the full
>software compatibility bit.
DIGITAL!!!!
wow.
Now wake me up when the first prototypes appear. (There has to be a
down side to everything :-) )
Ale.
Ale Terlevich
As were Apple when they decided to go for PowerPC.
Ale.
K.J. Bracey
> Digital and ARM Announce Agreement to Develop
> High-Speed, Low-Power, ARM-Compliant Processors
Yahoo! Whoop! Whoop! Whoop! <somersault> Yeehaaar! Ekee Ekee Ekee Zoooooooom!
Tom Hughes
It was originally posted in c.s.a.announce and c.s.arm, but somebody copied
it here for our information :-)
Tom
--
Tom Hughes (t...@compton.demon.co.uk)
5 Lampits, Hoddesdon, Herts EN11 8EH
... My hard disk is full! Maybe I'll try this message section thing.
Gavin Sallery
> kjb...@phx.cam.ac.uk (K.J. Bracey) writes:
>
> >In article <792031...@ivision.demon.co.uk> Ma...@ivision.demon.co.uk (Manar Hussain) writes:
>
> >> Digital and ARM Announce Agreement to Develop
> >> High-Speed, Low-Power, ARM-Compliant Processors
>
> >Yahoo! Whoop! Whoop! Whoop! <somersault> Yeehaaar! Ekee Ekee Ekee Zoooooooom!
>
Ditto. And more, with spatula-loads of excess delight and a large dollop of
KERRRPOWWWW!
> >I do believe that's the best news I've heard all year. Especially the full
> >software compatibility bit.
>
> >Kevin
> >=====
> Oh , YES!!!
[snip]
> Oh, ARM Ltd./Acorn come on! A simple 55MHz ARM 700 + FPA card will be
> sufficient for most of the users, and I don't think this is impossible
> at the moment since there were such cards shown on different shows!
> I think, the only reason for this deal is, that DEC has noticed, that their
> Alpha was a big fault (as the PowerPC, these CPUs have more than 200
> instructions each and call themselves RISC! The only advantages are bigger
> register sets (32 AFAIK), superscalar-architecure (the PowerPC can execute
> max. 3 instructions in one cycle), floating-point-instructions and higher
> clock-rates, but this costs them a lot of transistors and a lot of Watts - they
> are as hot as P90's! )
Uh... hang on, the thrust of your argument seemsto be... Acorns don't need to
be any faster? Que? What? I mean, doy? Yes. They're fast. They're very fast
in some areas, and having an OS that is slightly :-) less bloated than the
fat warthog that drives some of the PC clone offerings makes them seem
faster. But Intel & co. are advancing. AFAIK, a Pentium can outprocess a Risc
PC in flat performance terms (in some cases). This is annoying. The Alpha, on
the other hand, is simply *mindblowing*. You know the sort of transcendental
high you got when it was announced that Acorn were using something other than
the 6502 in their next generation machines... something than ran about 32
times faster? I was about... oh, I don't know, eight at the time, and I was
running up the walls having heard this. Manic. Completely high, just because
Acorn were using something that fast, faster than anyone else even knew how.
This is how I feel about the Alpha... it just kicks ass, thighs, calves,
heels and anything else you care to mention.
Yes, they're as hot as P90s... but much faster, and an entirely more elegant,
non-botched design. And with ARM's expertise in low-power applications...
we're talking something that runs on slightly more than ARM 8 current, and at
only slightly less than Alpha speeds. And compatible with ARM8. Baaahh!!!
Wow! Do you see now? This doesn't just do adequately, it's such a quantum
leap the rest of the computing industry will really be *nowhere* unless it
has something to compete with this thing. Well, that's how I feel right now.
It's fast.
And the wait won't be a problem... I need to start saving *NOW*!
--
______ _________
/__ (_ \ Hertfordshire, England
\_|avin __)allery \______________________
... Be *excellent* to each other
Kevin F. Quinn
nm...@minster.york.ac.uk (Bob) wrote:
It is...
--
Kevin F. Quinn * "That's not what you said when you sent him your
ke...@banana.demon.co.uk * Navel." "Novel, Baldrick, not navel."
ke...@cix.compulink.co.uk * "Well it sounds like a case of soggy grapefruits
Compu$erve: 100025,1525 * to me..." BlackAdder III
... I didn't know it was impossible when I did it.
Ollie Cornes
> this will be enough speed to run Windows'97's 16Bit code as fast as RiscOS-
> software on a A5k these days.
The word 'bollocks' floats freely in the front of my mind.
I doubt very much there will be any 16bit code anywhere near Windows 97.
Regards,
Ollie.
--
\ /\ Computer Concepts Ltd Xara Ltd
( ) ol...@cconcepts.co.uk ol...@xara.co.uk
.( o ). All opinions are mine
Mark Smith
ba...@cam.ac.uk (Ben Laughton) wrote:
> I heard a while ago that if current ARM's were thrown down the really
> modern processor manufacturing lines (does this imply a lower micron
> thingy?) then we would see an instant performance imporovement.
Hmm, I would have thought that GPS and VLSI have modern manufacturing lines.
What is the feature size of ARM chips currently in production? Somewhere
around the 0.6 micron mark, I should think, which is about the norm.
--
Mark Smith - Microelectronics Student & Acorn Software Developer.
All comments reflect my personal thoughts only. I reserve the right to be wrong.
.. Insanity doesn't just run in my family; it practically gallops
Michael Francis FI Sys Admin LDN
if we only add a faster processor? Is bus bandwidth not going to be a
problem ?
Mike
---
**********************************************************************
Mike Francis (francm) 'Any system failures are not my fault'
Disclaimer: All opinions are my own
**********************************************************************
S.J. Borrill
: > High-Speed, Low-Power, ARM-Compliant Processors
: Yahoo! Whoop! Whoop! Whoop! <somersault> Yeehaaar! Ekee Ekee Ekee Zoooooooom!
Nobody seems to have mentioned this, but.... I hope it will have an equally
fast floating point unit.
S.J.Borrill
Daniel Pead
es...@csv.warwick.ac.uk (Jon Ribbens) writes:
> Err, I think that the idea is that hopefully it will have implications
> for *Acorn* computers. (400MHz RISC PC anybody? Now wouldn't that be
> nice ;-) )
Yeah - plugged into a 15Mhz bus! Unless they have supernatural new
cache strategies, the words "returns" and "diminishing" spring to mind.
A bit like a Citroen 2CV GTI 4 litre turbo, really.
Ok, ok I did see the ":-)"...
Olly Betts
Michael Francis FI Sys Admin LDN <fra...@shula.fir.london> wrote:
>How much performance increase can we expect from the RISC PC architecture
>if we only add a faster processor? Is bus bandwidth not going to be a
>problem ?
PC clones seem to cope and their processor cores can run at 3 or 4 times the
speed. DEC Alphas can run at even higher rates.
An L2 cache (i.e. say 256K-1M of fast RAM between the processor and the bus)
helps a lot. It should be possible to put one onto a Risc PC processor
card, although it would probably be wasted at current clock rates.
Olly
--
Putting the "Ol" in technology.
Ben Laughton
Well, you're the electronics student so I stand corrected :-)
Ben
Tony Finch
> >
> > In article <3h5k4g$4...@lyra.csx.cam.ac.uk>,
> > ba...@cam.ac.uk (Ben Laughton) wrote:
> >
> > > I heard a while ago that if current ARM's were thrown down the really
> > > modern processor manufacturing lines (does this imply a lower micron
> > > thingy?) then we would see an instant performance imporovement.
> >
> > Hmm, I would have thought that GPS and VLSI have modern manufacturing lines.
> > What is the feature size of ARM chips currently in production? Somewhere
> > around the 0.6 micron mark, I should think, which is about the norm.
The StrongARMs will use a .35u process which is state of the art at the
moment -- that alone will give a big speed boost. Wahay!
BTW, does anyone else have trouble not being reminded of StrongEd?
Tony.
--
Coming to you from Long 0 7'3" E, Lat 52 12'21" N, Alt 8m.
Michael Williams
|[Alpha and PowerPC]
| The only advantages are bigger
|register sets (32 AFAIK), superscalar-architecure (the PowerPC can execute
|max. 3 instructions in one cycle), floating-point-instructions and higher
|clock-rates
Terrible, aren't they?
Mike. (Confused.)
____________________________________________________________________________
\ x / Michael Williams Advanced RISC Machines Limited
|\/|\/\ mwil...@armltd.co.uk Fulbourn Road, Cambridge, CB1 4JN, UK
| |(__)"I might well think that Matti, ARM Ltd. couldn't possibly comment."
Michael Williams
>Hmm, I would have thought that GPS and VLSI have modern manufacturing lines.
>What is the feature size of ARM chips currently in production? Somewhere
>around the 0.6 micron mark, I should think, which is about the norm.
I believe ARM610s are manufactured with a 0.8um feature size, and ARM710s
with a 0.6um feature size.
Mike. (Could be entirely wrong, of course.)
Tony Finch
> In article <1995Feb6.1...@arbi.Informatik.Uni-Oldenburg.DE>
Thilo....@arbi.informatik.uni-oldenburg.de (Thilo Manske) said:
>
> > this will be enough speed to run Windows'97's 16Bit code as fast as RiscOS-
> > software on a A5k these days.
>
> The word 'bollocks' floats freely in the front of my mind.
>
> I doubt very much there will be any 16bit code anywhere near Windows 97.
I doubt very much there won't be huge swathes of ancient, bloated 16 bit
code languishing in the filthy depths of Win95. Word, anyone? Excel? Win16
API? old device drivers?
Ale Terlevich
>In article <3h5k4g$4...@lyra.csx.cam.ac.uk>,
> ba...@cam.ac.uk (Ben Laughton) wrote:
>
>> I heard a while ago that if current ARM's were thrown down the really
>> modern processor manufacturing lines (does this imply a lower micron
>> thingy?) then we would see an instant performance imporovement.
>
>Hmm, I would have thought that GPS and VLSI have modern manufacturing lines.
>What is the feature size of ARM chips currently in production? Somewhere
>around the 0.6 micron mark, I should think, which is about the norm.
>
Acorn's processor roadmap puts the ARM8 at .6 to .35 microns.
Ale.
Ale Terlevich
>
>Uh... hang on, the thrust of your argument seemsto be... Acorns don't need to
>be any faster? Que? What? I mean, doy? Yes. They're fast. They're very fast
>in some areas, and having an OS that is slightly :-) less bloated than the
>fat warthog that drives some of the PC clone offerings makes them seem
>faster. But Intel & co. are advancing. AFAIK, a Pentium can outprocess a Risc
>PC in flat performance terms (in some cases). This is annoying. The Alpha, on
>the other hand, is simply *mindblowing*. You know the sort of transcendental
>high you got when it was announced that Acorn were using something other than
>the 6502 in their next generation machines... something than ran about 32
>times faster? I was about... oh, I don't know, eight at the time, and I was
>running up the walls having heard this. Manic. Completely high, just because
>Acorn were using something that fast, faster than anyone else even knew how.
>This is how I feel about the Alpha... it just kicks ass, thighs, calves,
>heels and anything else you care to mention.
Alphas are only 4 or 5 times faster than P90s.
>
>Yes, they're as hot as P90s... but much faster, and an entirely more elegant,
>non-botched design. And with ARM's expertise in low-power applications...
>we're talking something that runs on slightly more than ARM 8 current, and at
>only slightly less than Alpha speeds. And compatible with ARM8. Baaahh!!!
No we're not. If that were the case DEC would discontinue it's Alpha range.
If you read the their press release it states that the StrongARM will complement
the Alpha, not replace it. Alpha will continue to be the processor used for
high computing power applications. The only mention they make as to the speed
of this thing is to say that it won't be as powerful as an alpha.
>It's fast.
If you're saying a StrongARM is fast, how do you know? Have you seen one?
Do you have specs? Or are you just guessing.
Ale.
Martin Wuerthner
>>It's fast.
In the press releases it was said that the StrongARM would be about 3 to
4 times faster than the existing ARM610 and about the speed of the existing
PowerPC.
Martin
---------------------------------------------------------------------
Martin Wuerthner Computer Science and Linguistics
Crewe House
169, Mayfield Road m...@dcs.ed.ac.uk
Edinburgh, EH9 3AZ wuer...@minnie.informatik.uni-stuttgart.de
---------------------------------------------------------------------
Ben Laughton
> Alphas are only 4 or 5 times faster than P90s.
Oh sorry, is that it? :-)
Ben
S.W. Miller
No, that's not the whole story. They can also add up. :-)
Simon.
M J Ebourne
Gavin Sallery <ga...@sallery.demon.co.uk> wrote:
>Yes, they're as hot as P90s... but much faster, and an entirely more elegant,
>non-botched design. And with ARM's expertise in low-power applications...
>we're talking something that runs on slightly more than ARM 8 current, and at
>only slightly less than Alpha speeds. And compatible with ARM8. Baaahh!!!
Um, I think you need to learn a bit about processor design. Sure, ARM
design low power processors and DEC design fast ones. However, there's
very good reason why neither design low power fast processors. And
combining the two companies is rather easier than combining the two
chips.
Sure, they can design a pretty fast processor at relatively low power.
But there's a relationship between processor and electrical power -
the wrong way. Hence don't hold your breath for a 100mW 200 mips
processor.
Oh, and the processor speed depends somewhat on instruction set too.
Digital didn't pick an interesting looking instruction set and
make a fast processor of it - speed was built in from the start.
Various features of the ARM set cause problems (most noticeably
the conditional execution).
M J Ebourne
One is just assuming that all decent processors these days have fpu
ON CHIP. Hopefully...
Graham Allan
Huh? Find a faster memory bus than 15MHz? Not in many systems! Even the
233MHz Alpha machines use standard 70ns SIMMs (the new PCI-based systems,
anyway) - ie, about 15MHz. Of course, their memory bus is 64 bits wide
rather than 32, but the clock speed is the same.
Graham
Simon Proven
Thilo Manske <Thilo....@arbi.informatik.uni-oldenburg.de> wrote:
>We will get a machine that can compete with a 90MHz P5 in five years - in
>benchmarks, but Intel will have finished their work (and errata) with a
>200MHz P7 by this time - driving >>75 Watts ;-) at 2 Volts, >20M transistors,
>software on a A5k these days.
Hmm. You're out by about 4 years there:
From the F.T. 6 feb: "Digital will be in full scale production of
StrongARM chips in the first half of 1996"
--
-- Simon Proven : si...@camulos.demon.co.uk :
Rob Clark
> Well well, what good news. Nice to see there's life in the old dog
> yet :-)
> I heard a while ago that if current ARM's were thrown down the really
> modern processor manufacturing lines (does this imply a lower micron
> thingy?) then we would see an instant performance imporovement. (Is
> this because we'd be able to clock them faster - my knowledge in this
> area isn't great!).
I think so. The best Arm6's are about 40mhz at .6 micron. IBM (according
to PCW) have a .35 micron plant. The processors they make there are very
fast indeed. Fewer microns means more speed.
Good luck Arm/Digital, I was fearing Arm would be stuck to anonymity
but now maybe they'll find themselves beating Intel et al.
> Ben
> ba...@cam.ac.uk
--
__
Rob Clark ||
(_||_)
R.A....@herts.ac.uk (oo) Abe Lincoln's
RobC...@arcade.demon.co.uk /-------\/ cow
/ | ||
* ||----||
^^ ^^
Simon Proven
Mark Smith <ee91...@brunel.ac.uk> wrote:
>In article <3h5k4g$4...@lyra.csx.cam.ac.uk>,
> ba...@cam.ac.uk (Ben Laughton) wrote:
>> I heard a while ago that if current ARM's were thrown down the really
>> modern processor manufacturing lines (does this imply a lower micron
>> thingy?) then we would see an instant performance imporovement.
>Hmm, I would have thought that GPS and VLSI have modern manufacturing lines.
>What is the feature size of ARM chips currently in production? Somewhere
>around the 0.6 micron mark, I should think, which is about the norm.
State of the art fabs are about 0.35um at the moment; this is
considerably better than 0.6um. Of course there's more you
can do to an architecture to speed it up than reduce the
feature size; there have been plenty discussions on comp.sys.acorn
about these before.
Simon Proven
>if we only add a faster processor? Is bus bandwidth not going to be a
>problem ?
Depends how big the cache in StrongARM is, and on its cache
architecture, and on just how fast StrongARM is :-) A
second level cache could alleviate many of the problems.
Of course, a faster processor will always benifit from
a faster bus.
Mark Smith
rs...@unicorn.nott.ac.uk (Daniel Pead) wrote:
> Yeah - plugged into a 15Mhz bus! Unless they have supernatural new
> cache strategies, the words "returns" and "diminishing" spring to mind.
> A bit like a Citroen 2CV GTI 4 litre turbo, really.
Or even 16MHz.
<Pedant mode off>
--
Mark Smith - Microelectronics Student & Acorn Software Developer.
All comments reflect my personal thoughts only. I reserve the right to be wrong.
.. User - a technical term used by computer pros. See idiot.
Simon Truss
es...@csv.warwick.ac.uk (Jon Ribbens) wrote:
> for *Acorn* computers. (400MHz RISC PC anybody? Now wouldn't that be
> nice ;-) )
No, if I remember correctly the Alpha cannot be clocked at or over 300MHz
without becoming subject to SuperComputer export restrictions 8-(
but an Acorn computer that ran 5 times faster would be a dream, come true.
you could run several mpeg movies multitasking on the desktop in software.
I'll have to stop thinking about all this I'm too exited :-) :-) :-) :-) :-)
:-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-) :-)
Simon
--
mailto:si...@bigblue.demon.co.uk
mailto:u1...@csc.liv.ac.uk
http://www.csc.liv.ac.uk/users/u1smt/u1smt.html
... Feet Smell? Nose Run? Hey, you're upside down!
Robert J. Harley
>[...]
>a) Well all you guys shouting about why ARM don't push the performance
>envelope a bit - shut up ;)
Well the integer performance will be 3 or 4 times that of a Arm610,
so it will be similar to a PowerPC. But PowerPC has already been out for
ages and is moving forward too. There is no mention of FP performance.
The only envelopes being pushed are in low power, not performance.
Still, I'm not complaining! Unless they leave out the fast
multiplication instructions of the Arm7DM.
>b) Does anyone have a rough idea of when the new DEC FAB 6 in Hudson is
>due to open - that gives a strong clue as to when we will first see
>some StrongARMs...
Samples in second half of this year, according to Info World.
>c) I guess this explains where all those rumours about Newtons switching
>to PowerPC chips came from...
Don't know about that.
Bye,
Rob.
.-. rob...@vlsi.cs.caltech.edu .-.
/ \ .-. .-. / \
/ \ / \ .-. _ .-. / \ / \
/ \ / \ / \ / \ / \ / \ / \
/ \ / \ / `-' `-' \ / \ / \
\ / `-' `-' \ /
`-' Question all you are. `-'
Tony Finch
> In article <19950207....@mipc-08.brunel.ac.uk>
ee91...@brunel.ac.uk writes:
>
> I believe ARM610s are manufactured with a 0.8um feature size, and ARM710s
> with a 0.6um feature size.
ARM6 at 35MHz
ARM7 at 70MHz
StrongARM (.35um) at (guess) 200MHz??
Manar Hussain
m...@soton.ac.uk "M J Ebourne" writes:
> Sure, they can design a pretty fast processor at relatively low power.
> But there's a relationship between processor and electrical power -
> the wrong way. Hence don't hold your breath for a 100mW 200 mips
> processor.
The target 'wattage' of the StrongARM family is sub500mW -> 1W.
> Oh, and the processor speed depends somewhat on instruction set too.
> Digital didn't pick an interesting looking instruction set and
> make a fast processor of it - speed was built in from the start.
> Various features of the ARM set cause problems (most noticeably
> the conditional execution).
A thought just occurred to me (*pure speculation*): Maybe ARM were
hunting for the skills and resources of DEC to be able to get the
most out of their planned async processors?
Also whilst re-engineering the design to use say .35um fabrication
will certainly allow the chip to be clocked faster - will it not
allow more to be put on the chip. The prospect of being able to
throw on a fair bit more cache (up to 16K?) and an on-chip fpu and
then *only* upping the clock to 100MHz seems quite attractive.
BTW - one of the better likely consequences of all this (I hope)
is a really decent fpu, something DEC now a bit about...
> Martin Ebourne
Manar
--
Manar Hussain ma...@iVision.demon.co.uk
Graduate Horizons - http://www.ivision.co.uk/arcadia/horizons/
The W3 Careers Information Resource for the UK
contribtions welcome (eg career profiles) see TalkBack on site
Michael Williams
>In the press releases it was said that the StrongARM would be about 3 to
>4 times faster than the existing ARM610 and about the speed of the existing
>PowerPC.
In which press releases? I've checked the ones posted to comp.sys.arm and
can't find any such claims.
Mike.
Steve Turnbull
Is that all? Oh well might as well buy a PC then (only kidding, guys).
Steve Turnbull
(still) Editor, Acorn Computing (but just barely)
<really funny visual joke> from The Wrong Trousers
Torben AEgidius Mogensen
>Well well, what good news. Nice to see there's life in the old dog
>yet :-)
>I heard a while ago that if current ARM's were thrown down the really
>modern processor manufacturing lines (does this imply a lower micron
>thingy?) then we would see an instant performance imporovement. (Is
>this because we'd be able to clock them faster - my knowledge in this
>area isn't great!).
Basically, yes. If you go to a smaller feature size, you reduce the
capacitance of the connecting "wires", which allows signals to travel
and switch faster.
However, this is not the only way to speed things up. Things like
reducing the lengths of critical paths etc. are important too. I
suspect Digital has some tools for identifying critical paths and
possibly optimizing lay-out and connections. Don't expect 300MHz ARMs
straight away, though. ARM instructions do fairly much work which
can't all be parallelized (you can't do the shift in parallel with the
add in a shif-and-add instruction), so you can't push the clock rate
as far as one the Alpha, which was designed to minimize sequential
interdependencies inside and between instructions.
Torben Mogensen (tor...@diku.dk)
Thilo Manske
>In article <1995Feb6.1...@arbi.Informatik.Uni-Oldenburg.DE> Thilo....@arbi.informatik.uni-oldenburg.de (Thilo Manske) said:
>> this will be enough speed to run Windows'97's 16Bit code as fast as RiscOS-
>> software on a A5k these days.
>The word 'bollocks' floats freely in the front of my mind.
>I doubt very much there will be any 16bit code anywhere near Windows 97.
>Regards,
>Ollie.
>--
In Windows'95 (96, 97...?) there will be 16bit code:
Microsoft say, they want it to run in 4 MByte RAM, wich is not possible
with pure 32bit code... (i.e. their programmers are to lazy ;-) )
Why do you think, they will break this 'tradition' in the near future?
"If it's to slow, buy a faster machine" is the motto of nearly every
company that produces software for PCs... ;*}
EOF
--
------------------------------------------------------------------------------
Thilo....@Informatik.Uni-Oldenburg.DE
RiscPC600 ARM610 @33MHz 12+1MB RAM @16MHz 420MB IDE-HD RO3.50D
BBC A3000 ARM 2 @ 8MHz 4MB RAM @ 8MHz 250MB IDE-HD RO3.19 GU
I know which side of the soldering iron is the hot one...
*OUCH*
Sorri four may bed English.
Ollie Cornes
> but an Acorn computer that ran 5 times faster would be a dream, come true.
> you could run several mpeg movies multitasking on the desktop in software.
> I'll have to stop thinking about all this I'm too exited
Begs the question "why?".
Ollie.
--
\ /\ Computer Concepts Ltd Xara Ltd
( ) ol...@cconcepts.co.uk ol...@xara.co.uk
.( o ). All opinions are mine
Thilo Manske
>In article <1995Feb6.1...@arbi.Informatik.Uni-Oldenburg.DE> Thilo....@arbi.informatik.uni-oldenburg.de (Thilo Manske) wrote:
>> kjb...@phx.cam.ac.uk (K.J. Bracey) writes:
>>
>> >In article <792031...@ivision.demon.co.uk> Ma...@ivision.demon.co.uk (Manar Hussain) writes:
>>
>> >> Digital and ARM Announce Agreement to Develop
>> >> High-Speed, Low-Power, ARM-Compliant Processors
>>
>> >Yahoo! Whoop! Whoop! Whoop! <somersault> Yeehaaar! Ekee Ekee Ekee Zoooooooom!
>>
>Ditto. And more, with spatula-loads of excess delight and a large dollop of
>KERRRPOWWWW!
Do you really want CPUs which make noise? ;-]
>> >I do believe that's the best news I've heard all year. Especially the full
>> >software compatibility bit.
>>
>> >Kevin
>> >=====
>> Oh , YES!!!
>[snip]
Oh, you left the best part out ;-)
>> Oh, ARM Ltd./Acorn come on! A simple 55MHz ARM 700 + FPA card will be
>> sufficient for most of the users, and I don't think this is impossible
>> at the moment since there were such cards shown on different shows!
I just want to say, that I can't see a problem, why not build this CPU-Cards
today? And I want them!
Oh (too much Ohs ), there is a problem! Acorn garantied us prices of the
future CPU-Cards last summer, but they also promised an ARM800 in 'early
1995' and haven't even released an ARM700, or faster ARM600!
Why this CPU-Slots, when there aren't any faster CPUs available?
[snipping my own old text]
>Uh... hang on, the thrust of your argument seemsto be... Acorns don't need to
>be any faster? Que? What? I mean, doy?
*** NOT!!! *** Acorns need to go faster!! NOW!!!
Perhaps I've had better said 'will be sufficient at the moment', this had
made my position clearer...
I didn't want to say: ARM/Acorn stop making very fast CPUs for tomorrow.
I wanted to say: ARM/Acorn start making fast CPUs for today's use.
(i.e. e.g. 55Mhz ARM 700).
>....................................... Yes. They're fast. They're very fast
>in some areas,...
Yes, they're fast, but not very! You really can't do ray-tracing or any
other floating point thing on a RiscPC these days (of course, if you have
time enough, you can do ray-tracing even on a C-64, but there might be
memory problems...).
> ... and having an OS that is slightly :-) less bloated than the
>fat warthog that drives some of the PC clone offerings makes them seem
>faster. But Intel & co. are advancing. AFAIK, a Pentium can outprocess a Risc
>PC in flat performance terms (in some cases). This is annoying...
Sadly, a Pentium can outprecess a RiscPC in nearly all flat performance
terms, and the P6 will be released this year...
It's difficult, to find a benchmark, wich will make a RiscPC look faster
than a Pentium, simply because of the low clock-frequenzy.
> ... The Alpha, on
>the other hand, is simply *mindblowing*. You know the sort of transcendental
I know how fast they're. These benchmarks (benchmarks are useless, I know)
were figured out by me on our uni's Alphas (don't know what kind of Alpha's
they're):
FLOPS C Program (double Precision), V2.0 18 Dec 1992
[cut...................................................cut]
Iterations = 4194304 | You see something like this on my RiscPC:
NullTime (usec) = 0.0272 | | And this on a cheap 486DX2-66:
MFLOPS(1) = 16.4604 | 0.01 | 2-3
MFLOPS(2) = 15.5704 | 0.01 | 2-3 (all from mine
MFLOPS(3) = 22.7411 | 0.01 | 2-3 (human) memory)
MFLOPS(4) = 30.3082 | 0.01 | 2-3
(compiled with DEC OSF/1 AXP Compiler without any additional optimizing..)
Dhrystone Benchmark, Version 2.1 (Language: C)
Program compiled without 'register' attribute
Please give the number of runs through the benchmark: 1000000
[cut...................................................cut]
Microseconds for one run through Dhrystone: 7.8
Dhrystones per Second: 128479.7
(compiled as above)
[cut...................................................cut]
Microseconds for one run through Dhrystone: 4.8
Dhrystones per Second: 208333.3
(with additional -O4)
And you know how fast(?) a RiscPC is. (Acorn says 40K Dhrystones, I
measured something about 33K - a 486DX2-66 gets around 55K Dhrystones :-\
(with GCC -O2)).
But the Alpha is only fast because of 64-Bit data-paths and high clock-
rates - like the Pentium! The Pentium even is superscalar (though not very
efficient with 4 registers...), the Alpha 21064 can only perform integer
and floating point instructions parallely.
Where would we go, if ARM starts increasing clock-frequencies - a few
35 MHz ARM3s were available in the 90's even in Germany! But now, 5 years
later, I can't see any improvement. Ok, cached ARM 6xx have write buffers,
but the clock-rate is still low. - Very low - compared to 486 & PowerPC.
>high you got when it was announced that Acorn were using something other than
>the 6502 in their next generation machines... something than ran about 32
>times faster? I was about... oh, I don't know, eight at the time, and I was
>running up the walls having heard this. Manic. Completely high, just because
>Acorn were using something that fast, faster than anyone else even knew how.
Yes, there was a quantum leap, but since then not very much has happend :-(
1987 when the Archimedes came out, it was in front of all the other
computers for the 'little man' (I don't want to use the words 'Personal
Computer', I ment such things as Amiga, Atari STs, 286'ers...)
Today, the RiscPC is 5-6 times faster as the old Archimedes, but look at the PCs?
The standard PCs of 87 (286'er without Copro, I believe) were *VERY* slow (compared
to the Archimedes) but now?
>This is how I feel about the Alpha... it just kicks ass, thighs, calves,
>heels and anything else you care to mention.
>Yes, they're as hot as P90s... but much faster, and an entirely more elegant,
>non-botched design. And with ARM's expertise in low-power applications...
Of course they're faster and better. Nearly every CPU is better (not in
performance) than a Pentium (four specialised registers ***urgh***! ). And
even a 6502 has a more elegant design ;-). Pentium - seems to me, that Intel
tries to go supersonic with a steam-boat.
But you can buy a complete Pentium-PC with 8 MB to the same price as a
simple RiscPC 600-5. Of course, its crap, nearly unusably (does this word
exist?) unstable, but its faster :-(.
>we're talking something that runs on slightly more than ARM 8 current, and at
>only slightly less than Alpha speeds. And compatible with ARM8. Baaahh!!!
Fine! It's compatible with ARM8. Who has an ARM8xx? Who has an ARM7xx???
(ARM ltd. has, haha ;-) We all have only have ARM610s in our RiscPCs. Ok,
there are lucky ones, who get those 55MHz-ARM700 card prototypes, but the
rest of us only has 30MHz ARM610s. (I've tuned mine to run @33MHz ;-),
AFAIK the ARM610 is able to run even @40Mhz, but the additional chips and
resistors on the newer 610-cards (as mine) don't like this speed :-( )
What I want is a fast ARM-CPU these days, not years in the future.
And with the ability to drive a FPA. The only reason (that makes sense to me)
for Acorn to release the RiscPC with ARM610s not ARM600s is that they're
very cheap, because Apple has needed them in greater numbers for the Newtons,
so the producing cost dropped.
What do you mean with 'Baaahh!!!'? (English is knot my mother-tongue, you know...)
>Wow! Do you see now? This doesn't just do adequately, it's such a quantum
>leap the rest of the computing industry will really be *nowhere* unless it
>has something to compete with this thing. Well, that's how I feel right now.
Even a small step (from 30 to >50MHz and FPA) will make me happy this year...
>It's fast.
Let's hope the P7 won't be faster.
>And the wait won't be a problem... I need to start saving *NOW*!
Though I'm only a student, I've got the money, and I want a fast CPU
with FPA **NOW** The ARM-700 was promised to come out in the last quarter
of 1994 and now I've heard it will be released this summer (sounds like
Windows'95, funny, isn't it?)
Bye. Thilo, a proud RiscPC-owner, who wants it to go faster...
Matthew Hambley
--
In article <792031...@ivision.demon.co.uk>, Ma...@ivision.demon.co.uk (Manar
Hussain) writes:
|>Embargoed Until Monday February 6, 1995 UK62
|>
|><hmm - didn't stop the Sunday Times... - ed>
|>
|> Digital and ARM Announce Agreement to Develop
|> High-Speed, Low-Power, ARM-Compliant Processors
|>
Drool, drool, gibber, gibber. There go the old kness. <Colapses to
ground prostrate and cries> I'm not worthy.
When will the Strong RISC PC be coming out? Acorn start developing
*NOW* please.
(I am quite incoherent with joy!)
-------------------------------------------------------------------------
Matthew Hambly | "It doesn't matter if the cat is black
cee...@caledonia.hw.ac.uk | or white as long as it's been cooked
http://www.cee.hw.ac.uk/~ceemah | in a Wok." - K. Thomas
-------------------------------------------------------------------------
Tim Gladding
94...@eng.cam.ac.uk (S.W. Miller) wrote:
So can P90's. Its fp division they occasionally have a problem with.
>
>
>
>
>
>
>
>
>
>
>
>
^^ blank lines.
Eek!
--
Tim Creator of alt.fan.piers and proud of it :-)
Cambridge, England. Try it, for all your Piers Wombwell needs!
Ollie Cornes
> I doubt very much there won't be huge swathes of ancient, bloated 16 bit
> code languishing in the filthy depths of Win95. Word, anyone? Excel? Win16
> API? old device drivers?
Office for NT has just been released. 32 bit versions of Word & Excel
included.
I s'pose it's more likely that Win 97 will in fact be NT 97.
The underlying basis of Windows 95 isn't exactly the best in the
world.. It seems fair that Microsoft are using it as a plank
to take you over the river to NT.
Just for those that don't know. NT doesn't run on DOS..
Regards,
Mark Smith
fa...@cam.ac.uk (Tony Finch) wrote:
> In article <3h7snh$9...@doc.armltd.co.uk>, mi...@nsict.org wrote:
>
> > In article <19950207....@mipc-08.brunel.ac.uk>
> ee91...@brunel.ac.uk writes:
> >
> > I believe ARM610s are manufactured with a 0.8um feature size, and ARM710s
> > with a 0.6um feature size.
Err, I think that was Mike, not me!
--
Mark Smith - Microelectronics Student & Acorn Software Developer.
All comments reflect my personal thoughts only. I reserve the right to be wrong.
.. One person's <grin> is another's <groan>.
Philip Sainty
>non-botched design. And with ARM's expertise in low-power applications...
>only slightly less than Alpha speeds. And compatible with ARM8. Baaahh!!!
>leap the rest of the computing industry will really be *nowhere* unless it
>has something to compete with this thing. Well, that's how I feel right now.
"Quantum" : A minimum amount of a physical quantity which can exist and by
multiples of which changes in the quantity occur.
I don't for a moment imagine that the StrongARM will be running at "only
slightly less than Alpha speeds". You say yourself that the Alpha's speed
is due to its design, and the StrongARM is to be compatible with the ARM
architecture, not the Alpha.
I am very excited by the possibilities of the StrongARM, however, and I
hope that having Digital in the picture will mean that the project will
see light-of-day in the not too distant future...
Philip
Adrian Bool
> Figured this would be of interest..
>
<cut>
> Digital and ARM Announce Agreement to Develop
> High-Speed, Low-Power, ARM-Compliant Processors
>
<cut>
> Performance is Key
>
> According to Robin Saxby, Managing Director & CEO of ARM: "Having...
<cut>
> ...The agreement with Digital will maintain our lead in these areas
> while allowing us to pursue applications demanding very high
> absolute performance."
Does this mean that ARM are leaving Digital to continue with the
design and manufacture of low power ARMs whilst ARM themselves
start concentrating on making REALLY fast chips - ie. absolute
performance ?
This seem like one of the most interseting bits of the announcement...
Adrian Bool
Whey hey! An arm at Pentium speeds at least! SHouldn't be too hard
for them since they do not need to woory about all the
8086 rubbish that INtel suffer from!
Ale Terlevich
In article r...@doc.armltd.co.uk, michael....@armltd.co.uk (Michael Williams) writes:
>In article <mxw.79...@dcs.ed.ac.uk> m...@dcs.ed.ac.uk (Martin Wuerthner) writes:
>>In the press releases it was said that the StrongARM would be about 3 to
>>4 times faster than the existing ARM610 and about the speed of the existing
>>PowerPC.
>
>In which press releases? I've checked the ones posted to comp.sys.arm and
>can't find any such claims.
>
for the Arm8 is meant to me about 3-4 times faster than the ARM6. Perhaps Martin
was getting confused with this?
Ale.
Ale Terlevich
In article F...@fi.gs.com, fra...@shula.fir.london (Michael Francis FI Sys Admin LDN) writes:
>How much performance increase can we expect from the RISC PC architecture
>if we only add a faster processor? Is bus bandwidth not going to be a
>problem ?
>
That depends how you design the processor. If you stuck an ultra fast processor
with 64MB of ultra fast ram on the processor card, then it would only need to
use the bus for IO, and 64MB/s should be fast enoung for that :-)
Ale.
Robert Templeman
> How much performance increase can we expect from the RISC PC architecture
> if we only add a faster processor? Is bus bandwidth not going to be a
> problem ?
> Mike
well, how about at least 100% more!. I've been writing texture
mapping routines for the Arc and i'm utterly horrified at how damn
fast my A410 is compared to the RPC600 (especially since my 410 is
essentially 6-7 years old, compare a 386 to a pentium)
My 410 has an 8Mhz bus which gets flooged with video refresh too.
I've found that every clock speed increase i've tried makes a
noticable difference in the usable feel of the machine. Moving from
ARM2 TO ARM3 is awesome at 25Mhz its at least 3 times faster at most
things. moving to 36Mhz and then 42 which is what i've stuck with,
the machine is noticably faster than a 25Mhz machine.
The RPC has a 16Mhz main bus which hasnt got to cope with Video DMA
so effectively its got 3X the bandwidth of my 410 so i reckon the
performance vs cache speed would probably only start to fold
significantly when in excess of 100MHz. Ive seen projected Dhrystone
benchmark figures in a graph in an RPC flyer which shows tail off at
much lower clock speeds, i think usability, especially with graphics
would continue to benefit at much higher clock rates.
Robert Templeman
> Alphas are only 4 or 5 times faster than P90s.
> >
> >Yes, they're as hot as P90s... but much faster, and an entirely more elegant,
Ahhemm, run a tight loop sure, but do anything bus-floggy and poor mr
Pentium dies a very horrible death, very quickly.
A beautiful benchmark is to diagonalise a 1000x1000 array, i havent
as yet been able to get any pentium to do that! All the Alpha
machines managed it with flying colours.
--
"I feel like a one legged cat burying turds on a frozen pond" - Mr Anderson
Clive...@armltd.co.uk
>> ...The agreement with Digital will maintain our lead in these areas
>> while allowing us to pursue applications demanding very high
>> absolute performance."
>
>Does this mean that ARM are leaving Digital to continue with the
>design and manufacture of low power ARMs whilst ARM themselves
>start concentrating on making REALLY fast chips - ie. absolute
>performance ?
I think you need to read the quote again, a little more carefully. I
read it as saying that the Digital deal allows us to pursue
applications demanding very high absolute performance, and it should
be fairly clear that StrongARM is our way of pursuing them.
--Clive.
(Disclaimer: I wouldn't believe a word of this if I were you...)
Pieter L. Molenaar
>K.J. Bracey (kjb...@phx.cam.ac.uk) wrote:
>: In article <792031...@ivision.demon.co.uk> Ma...@ivision.demon.co.uk (Manar Hussain) writes:
>: > Digital and ARM Announce Agreement to Develop
>: > High-Speed, Low-Power, ARM-Compliant Processors
[stuff eaten]
Anyone remember the information sheets for the RiscPC when it came out
in May ? I recall the sheet with the upgrade-prices for the new processors,
and I noticed that besides the Arm8 there was an 'high-performance Arm8'.
Now I think Acorn has been naughty to us and not told us about
the DEC/Arm-deal.
Just a thought :=}.
Pieter,
pmol...@fwi.uva.nl
--
================= ================ ==================
Pieter Molenaar login:pmol...@fwi.uva.nl
UvA (BIS 92)
================= ================ ==================
Wayne Bagguley
In article <D3Mqz...@festival.ed.ac.uk>, a...@roe.ac.uk (Ale Terlevich)
writes:
[snip]
|> >Yes, they're as hot as P90s... but much faster, and an entirely more
|> elegant,
|> >non-botched design. And with ARM's expertise in low-power applications...
|> >we're talking something that runs on slightly more than ARM 8 current, and
|> at
|> >only slightly less than Alpha speeds. And compatible with ARM8. Baaahh!!!
|>
|> No we're not. If that were the case DEC would discontinue it's Alpha
|> range.
|> If you read the their press release it states that the StrongARM will
|> complement
|> the Alpha, not replace it. Alpha will continue to be the processor used for
|> high computing power applications. The only mention they make as to the
|> speed
|> of this thing is to say that it won't be as powerful as an alpha.
I'm sorry, but why would DEC discontinue their Alpha range, the ARM is totally
different,
people who wanted an Alpha would buy one, and people who wanted an StrongARM
would buy one of
_them_.
And why would it not be as powerful as the Alpha, that's like saying the ARM8
won't be as
powerful as the ARM7!
In any case, just because they are designing the StrongARM doesn't mean that
they have
stopped doing everything else.
They're propably gonna release a new superfast Alpha next year anyway.
|>
|> >It's fast.
True.
|>
|> If you're saying a StrongARM is fast, how do you know? Have you seen
|> one?
|> Do you have specs? Or are you just guessing.
|>
|> Ale.
|>
|>
They're not exactly going to be slow are they, come on.
"Oh let's get together with ARM and make a really slow processor that no-one
wants."
Alternatively.
"Oh let's get together with DEC and make a really slow processor that no-one
wants."
Be real.
Think before you write.
Ta ta.
Wayne :)
Philip Colmer
According to a trade magazine (Electronics Weekly, I think), they reckoned
that the ARM8 was supposed to be doing 60MIPs and the StrongARM would be
doing 200MIPs. A reasonable chunk of the article was spent speculating as to
how the processor would achieve that speed - part of it boiled down to
getting the clock per instruction rate down to 1 and then clocking the
processor at 200MHz and another part involved using the 0.35um process.
All of this is taken from the article and no other source, so don't blame me
if its wrong :-)
--Philip
---------------------------------------------------------------------
Philip Colmer, Senior Systems Engineer
Acorn Computers Ltd, Acorn House, Vision Park, Histon, Cambridge,
England, CB4 4AE.
Tel: 01223 254254 Fax: 01224 254262
The views expressed in this article are those of the author and not
necessarily those of Acorn Computers Ltd.
Richard West
|> In article <19950207....@bigblue.demon.co.uk> Si...@bigblue.demon.co.uk (Simon Truss) said:
|>
|> > but an Acorn computer that ran 5 times faster would be a dream, come true.
|> > you could run several mpeg movies multitasking on the desktop in software.
|> > I'll have to stop thinking about all this I'm too exited
|>
|> Begs the question "why?".
|>
Guess that depends on what sort of movies he's watching...
Thilo Manske
>In article <1995Feb6.1...@arbi.Informatik.Uni-Oldenburg.DE>,
>Thilo Manske <Thilo....@arbi.informatik.uni-oldenburg.de> wrote:
>>We will get a machine that can compete with a 90MHz P5 in five years - in
[snip]
>Hmm. You're out by about 4 years there:
>From the F.T. 6 feb: "Digital will be in full scale production of
> StrongARM chips in the first half of 1996"
>--
>-- Simon Proven : si...@camulos.demon.co.uk :
Yes, this is what they **SAY**.
But do you really believe, you can go to your local dealer in summer 1996
and buy a StrongARM-Card for your RiscPC? I don't think so, look what
happens to the PC-Card, ARM700 or ARM800 cards? You should be able to buy
them all today, but in Germany not even the PC-Card has been released to
trade.
I said we will get a working MACHINE in five years, not a CPU.
There's a difference, you know.
Graham Allan
> A thought just occurred to me (*pure speculation*): Maybe ARM were
> hunting for the skills and resources of DEC to be able to get the
> most out of their planned async processors?
Here's another bit of very wild speculation then: maybe Digital are
looking more closely at the set-top box market (via Online Media?).
After all, they are making a big effort in the server end of the same
market. Could that be the plan: Alphas for the media servers (where the
64-bit address space & filesystem is genuinely useful); ARMs for the
clients? Well, this is probably reading too much into it, but it's an
interesting coincidence.
Graham
F.A.N. Finch
<bagg...@cs.man.ac.uk> wrote:
>
>And why would it not be as powerful as the Alpha, that's
>like saying the ARM8 won't be as powerful as the ARM7!
Reasons why the StrongARM will be slower than the Alpha:
1. The Alpha's clock runs probably 2x faster
2. The Alpha is a 64 bit machine as opposed to 32 bits
3. Tha Alpha has more registers (32 rather than 16)
4. The Alpha is superscalar (it can complete more than one
instruction per cycle).
5. The Alpha has 100x the transistor budget and power budget
6. The Alpha's instruction set is designed for speed in a large cached
processor. The ARM's is designed to make the most out of a small die
and no cache.
OK?
Tony.
--
Coming to you from Long 0 7'ish E, Lat 52 12'ish N, Alt 8m.
I relly must work out the proper position of this place...
--
Ale Terlevich
>In article <D3Mqz...@festival.ed.ac.uk>, a...@roe.ac.uk (Ale Terlevich)
>writes:
>
>[snip]
>
>|> >Yes, they're as hot as P90s... but much faster, and an entirely more
>|> elegant,
>|> >non-botched design. And with ARM's expertise in low-power applications...
>|> >we're talking something that runs on slightly more than ARM 8 current, and
>|> at
>|> >only slightly less than Alpha speeds. And compatible with ARM8. Baaahh!!!
>|>
>|> No we're not. If that were the case DEC would discontinue it's Alpha
>|> range.
>|> If you read the their press release it states that the StrongARM will
>|> complement
>|> the Alpha, not replace it. Alpha will continue to be the processor used for
>|> high computing power applications. The only mention they make as to the
>|> speed
>|> of this thing is to say that it won't be as powerful as an alpha.
>
>I'm sorry, but why would DEC discontinue their Alpha range, the ARM is totally
>different,
>people who wanted an Alpha would buy one, and people who wanted an StrongARM
>would buy one of
>_them_.
I never said they would! There is *no way* Digital will discontinue it's
alpha range, because there is *no way* the SrongARM will be as fast!
After all, why would DEC make a chip that would compete with the Alpha?
>And why would it not be as powerful as the Alpha, that's like saying the ARM8
>won't be as
>powerful as the ARM7!
No it's not! The ARM8 will be a progression from the ARM7. The StrongARM
will not be a progression from the Alpha.
The reason the StrongARM won't be as powerful as the Alpha, is that it
is constrained to *low power consumption* whereas the Alpha isn't. These
two chips have different aims. The aim of the alpha is sheer speed.
Furthermore, for some applications the ARM6 in my RiscPC is slower than
the Arm3 in my A310....
>In any case, just because they are designing the StrongARM doesn't mean that
>they have
>stopped doing everything else.
Of course not.
>
>They're propably gonna release a new superfast Alpha next year anyway.
Yep.
>
>|>
>|> >It's fast.
>
>True.
>|>
>|> If you're saying a StrongARM is fast, how do you know? Have you seen
>|> one?
>|> Do you have specs? Or are you just guessing.
>|>
>They're not exactly going to be slow are they, come on.
>"Oh let's get together with ARM and make a really slow processor that no-one
>wants."
>
>Alternatively.
>
>"Oh let's get together with DEC and make a really slow processor that no-one
>wants."
>
>Be real.
>
>Think before you write.
The point is that a lot of people keep going on about how fast this thing's
going to be, and none of you have any idea what you're talking about. The logic
I was arguing against, seems to go like this,
DEC make Alphas
DEC are partners in StrongARM
Therefore StrongARM will be nearly as fast as the Alpha.
I'm not saying that they're going to be slow, However when compared to
other chips that are going to come out in a year's time, I don't think they'll
be particularly fast. Why should they be? After all their selling point will
be lower power consumption.
Ale.
Ale Terlevich
In article d...@rockall.cc.strath.ac.uk, cad...@ccsun.strath.ac.uk (Kenny MacLeod) writes:
>In article r...@acorn.acorn.co.uk, pco...@acorn.co.uk (Philip Colmer) writes:
>>According to a trade magazine (Electronics Weekly, I think), they reckoned
>>that the ARM8 was supposed to be doing 60MIPs and the StrongARM would be
>>doing 200MIPs. A reasonable chunk of the article was spent speculating as to
>>how the processor would achieve that speed - part of it boiled down to
>>getting the clock per instruction rate down to 1 and then clocking the
>>processor at 200MHz and another part involved using the 0.35um process.
>
>fingers at anyone, Philip ;-) ) starts quoting MIPS figures?
>
Not for two processors with the same instruction set!!
200MIPs eh?
WOW! ~10 times faster than an ARM6!
Ale.
Kenny MacLeod
Unlike, of course, the 21164, which really crawls along ;-)
---
Cheers
K
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Kenny MacLeod cad...@ccsun.strath.ac.uk
http://www.strath.ac.uk/~cadp59/homepage.html
Computer & Electronic Systems
University of Strathclyde
Kenny MacLeod
>that the ARM8 was supposed to be doing 60MIPs and the StrongARM would be
>doing 200MIPs. A reasonable chunk of the article was spent speculating as to
>how the processor would achieve that speed - part of it boiled down to
>getting the clock per instruction rate down to 1 and then clocking the
>processor at 200MHz and another part involved using the 0.35um process.
Grr. Am I the only one who's hair stands on end when someone (pointing no
fingers at anyone, Philip ;-) ) starts quoting MIPS figures?
---
Richard Birkby
In article <3hddlo$d...@rockall.cc.strath.ac.uk>, cad...@ccsun.strath.ac.uk (Kenny MacLeod) writes:
|> In article r...@acorn.acorn.co.uk, pco...@acorn.co.uk (Philip Colmer) writes:
|> >According to a trade magazine (Electronics Weekly, I think), they reckoned
|> >that the ARM8 was supposed to be doing 60MIPs and the StrongARM would be
|> >doing 200MIPs. A reasonable chunk of the article was spent speculating as to
|> >how the processor would achieve that speed - part of it boiled down to
|> >getting the clock per instruction rate down to 1 and then clocking the
|> >processor at 200MHz and another part involved using the 0.35um process.
|>
|> Grr. Am I the only one who's hair stands on end when someone (pointing no
|> fingers at anyone, Philip ;-) ) starts quoting MIPS figures?
Not in this context as its referring to the same processor. The same reason
applies for always quoting Mhz on ARM processors. The figures are always
going to be slightly out due to changes in the architecture, but close enough
to give a very good idea.
Richard
Christopher Paul Rijk
If think a good description of what DEC are going to consider when
designing various chips can be boiled down to:
StrongARM family
How fast can we make it, keeping it compatible (with other ARM's)
having power consumption less than 1W, price less than 100 USD in bulk.
(figures just my guesses)
Alpha family
How fast can we make it, keeping it compatible (with other Alpha's)
and have it available at a marketable price.
I'm sure we can all see the difference.
--
+-- Chris Rijk --------------------------------------- cp...@doc.ic.ac.uk --+
| Dictionary of Alternative Meaings: nation |
| A society united by a delusion about it's ancestry and by a common |
| hatred of it's neighbours |
Gavin Sallery
[Over-excited dribblings from me deleted]
> Um, I think you need to learn a bit about processor design. Sure, ARM
> design low power processors and DEC design fast ones. However, there's
> very good reason why neither design low power fast processors. And
> combining the two companies is rather easier than combining the two
> chips.
I know.
> Sure, they can design a pretty fast processor at relatively low power.
> But there's a relationship between processor and electrical power -
> the wrong way.
I know.
> Oh, and the processor speed depends somewhat on instruction set too.
You're kidding?
> Digital didn't pick an interesting looking instruction set and
> make a fast processor of it - speed was built in from the start.
Yes.
OK, point taken. Fortunately, I wasn't actually being *totally* serious at
the time; you may have noticed that this thread is prone to a little hysteria
at the moment. Large increases in speed can do that to people. Unfortunately,
you found it necessary to lecture me on processor design. Shame I got a
little carried away, but I was talking in idealistic terms; it seemed a good
day for it.
And besides, I didn't mention a timescale :-)
--
______ _________
/__ (_ \ Hertfordshire, England
\_|avin __)allery \______________________
... It all comes down to treating others as you want to be treated.
Keith Marlow
PC>According to a trade magazine (Electronics Weekly, I think), they
PC>reckoned
PC>that the ARM8 was supposed to be doing 60MIPs and the StrongARM would
PC>be
PC>doing 200MIPs. A reasonable chunk of the article was spent speculating
PC>as to
PC>how the processor would achieve that speed - part of it boiled down to
PC>getting the clock per instruction rate down to 1 and then clocking the
PC>processor at 200MHz and another part involved using the 0.35um process
PC>
PC>
Something that doesn't seem to have been discussed is the final price of the
StrongARM? Using a 0.35um process must push the final price up rather a lot,
cutting edge never comes cheap.
Just my ~2 pence worth (I've been using a Pentium, so I can't be exact)
Keith
Martin Wuerthner
Well, I was referring to the following report:
REPORT IN INFO WORLD 6-2-95
[... stuff deleted ...]
The two companies said they plan to make the StrongARM processors twice as
powerful as other processors that use less than half a watt of power.
Analysts briefed on the announcement said the StrongARM processor, which
^^^^^^^^^^^^^^^^^^^^^^^
will be an upgrade to the upcoming ARM810 chip, will be three to four times
^^^^^^^^^^^^^^^^^^^^^^^^^^^
as powerful as the ARM610 chip, or similar to existing PowerPC processors.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Martin
---------------------------------------------------------------------
Martin Wuerthner Computer Science and Linguistics
Crewe House
169, Mayfield Road m...@dcs.ed.ac.uk
Edinburgh, EH9 3AZ wuer...@minnie.informatik.uni-stuttgart.de
---------------------------------------------------------------------
Ben Laughton
>
> Snip...
>
> I s'pose it's more likely that Win 97 will in fact be NT 97.
> The underlying basis of Windows 95 isn't exactly the best in the
> world.. It seems fair that Microsoft are using it as a plank
> to take you over the river to NT.
>
> Just for those that don't know. NT doesn't run on DOS..
>
Indeed, isn't the general idea to gradually aim everyone twoards
their Cairo system (I do get confused with all these names, so I
maybe wrong). In the meantime, is it Daytona that is Windows NT with
a Windows 95 style interface?
Ben
Dave Cox
[ Slash ]
: Huh? Find a faster memory bus than 15MHz? Not in many systems! Even the
: 233MHz Alpha machines use standard 70ns SIMMs (the new PCI-based systems,
: anyway) - ie, about 15MHz. Of course, their memory bus is 64 bits wide
: rather than 32, but the clock speed is the same.
I just read that Apple are planning to ship machines with PCI buses in the
Autum sometime. They also have a new set of Power Macs. Top notch is the
PowerMac 8100/110, seriously nifty piece of kit at US$6895 roughly. Now
if only they'd put a decent OS on it...
Oh, while we are on buses, the 8100/110 has a 40mhz 64bit data bus I
believe..Damn...
No wonder it thunders through MPEG movies etc...
--
Dave Cox email: cm2b...@bs47c.staffs.ac.uk
Torben AEgidius Mogensen
>How much performance increase can we expect from the RISC PC architecture
>if we only add a faster processor? Is bus bandwidth not going to be a
>problem ?
It will certainly be a limiting factor. How much so depends on the
applications and on the cache size/structure. In short, it depends on
the hitrate of the cache and on the relative costs of cache/memory
access. Below I have a table of relative performances of processors
running at various speeds, compared to a processor running without
cache at memory speed. A hit costs one cache cycle and a miss costs
one cache cycle plus one bus cycle. This means that a processor
running with cache might be slower than a processor of the same speed
without cache.
cache speed relative speed at various hit rates
(x bus speed) 70% 80% 90% 95% 97%
1 0.77 0.83 0.91 0.95 0.97
2 1.25 1.43 1.67 1.82 1.89
3 1.58 1.88 2.31 2.61 2.75
4 1.82 2.22 2.86 3.33 3.57
5 2.00 2.50 3.33 4.00 4.35
6 2.14 2.73 3.75 4.62 5.08
7 2.25 2.92 4.12 5.19 5.79
8 2.35 3.08 4.44 5.71 6.45
infinite 3.33 5.00 10.00 20.00 33.00
If you double the cache, you will (by a VERY rough estimate) go one
column to the right due to increased hit rate (this is probably a bit
optimistic). The hit rate is also affected by cache structure,
e.g. write-back/write-through, unified/separate and whether or not
there is a write-buffer. If you assume improvements in cache structure
when the size increases, the one column per doubling estimate is
probably not far off.
The "infinite" line represents the limit of performance at various hit
rates. It shows that at low hit rates you can't get very fast, but
that there is plenty of room for improvement if you have a high hit
rate.
Presently the cache on the RiscPC runs at about twice the speed of the
bus, so it corresponds roughly to the second line. A 66MHz ARM7x0 will
correspond to line 4, but one column to the right due to larger
cache. It will thus make the RiscPC run almost twice as fast, unless
the applications are very memory intensive (in which case the hit rate
can be far below 70%). If you have a 128MHz (Strong)ARM with an even
larger cache, you end up in line 8, shifting two columns right
compared to the present RiscPC, thus getting almost four times the
speed.
Torben Mogensen (tor...@diku.dk)
Adam Goodfellow
Ben Laughton <ba...@cam.ac.uk> wrote:
>a Windows 95 style interface? ^^^^^^^
> ^^^^^^^
^^^^^^^
Sort of appropriate for something that crashes alot ;-)
--
Adam
=======================================================================
| Computech Tel/Fax: 0181 673 7817 email: ad...@comptech.demon.co.uk |
=======================================================================
Philip Sainty
>And why would it not be as powerful as the Alpha, that's like saying the ARM8
>won't be as powerful as the ARM7!
That's nothing like saying that the ARM8 won't be as powerful as the ARM7.
The ARM8 is a advanced version of the ARM7.
The StrongARM will *not* be an advanced version of the Alpha.
>Think before you write.
Philip
Thorsten Seitz
> [...]
>Depends how big the cache in StrongARM is, and on its cache
>architecture, and on just how fast StrongARM is :-) A
>second level cache could alleviate many of the problems.
> [...]
Transfer-rates in second level caches of modern Pentium boards
(66MHz/90MHz) are around 40MB/s, as measured by the renowned german
computer-magazine c't.
That's a bandwidth of 80MB/s -- not much faster than the main memory
of a Risc PC, which should reach about 64MB/s (Does it really? What
about the N-cycles? Anybody measured the transfer-rate (simple
load-store-loop)?)
-Thorsten
Tim Browse
<sigh> Why is it whenever we hear predictions like that it always
turns out to be about 10-30% faster in real terms? :-(. No matter
who made the CPU (DEC, Acorn, Intel, Gould etc)
The only exception I can think of is when the ARM arrived. Would
be nice to see that performance leap again, but I'm just too
cynical to believe it...maybe we should all just live a pastoral
existence with no machines to hassle us? :-)
ARM2->ARM3 was quite nice though...but I upgraded from RO2 to RO3
at the same time and that seemed to soak up a lot of that lovely
CPU time I'd just bought. Oh well.
Tim Browse
Software Engineer, Computer Concepts Ltd/Xara Ltd
t...@cconcepts.co.uk/t...@xara.co.uk
"This stuff we do...it isn't ready yet."
Ian Griffiths
: benchmarks, but Intel will have finished their work (and errata) with a
: 200MHz P7 by this time - driving >>75 Watts ;-) at 2 Volts,
: ^^^^^^^
That would be a good trick.
--
Ian Griffiths | "There's no such thing as a humble opinion" -
i...@intrac.demon.co.uk | (Captain Vimes, Men at Arms by Terry Pratchett)
Ian Griffiths
: es...@csv.warwick.ac.uk (Jon Ribbens) writes:
: > Err, I think that the idea is that hopefully it will have implications
: > for *Acorn* computers. (400MHz RISC PC anybody? Now wouldn't that be
: > nice ;-) )
: Yeah - plugged into a 15Mhz bus! Unless they have supernatural new
: cache strategies, the words "returns" and "diminishing" spring to mind.
: A bit like a Citroen 2CV GTI 4 litre turbo, really.
Oh I don't know. 2CVs don't suffer from diminishing returns in the
same way. A 4 litre Turbo 2CV would go very very fast indeed.
The only problem would be an inability to see over the bonnet, with
an engine that large, and probably an inability to go in straight
lines, since the road holding of four bicycle tires is not great evenm
with a pathetic engine, let along something nippy.
Ian Griffiths
: Alphas are only 4 or 5 times faster than P90s.
You sure? The latest Alphas run at about 1300 MIPS. That would make
a P90 about 260 MIPS. I don't believe I get that out of mine.
Ian Griffiths
: In article <19950206....@cconcepts.co.uk>, ol...@cconcepts.co.uk wrote:
: > I doubt very much there will be any 16bit code anywhere near Windows 97.
: I doubt very much there won't be huge swathes of ancient, bloated 16 bit
: code languishing in the filthy depths of Win95. Word, anyone? Excel? Win16
: API? old device drivers?
Word, Excel, NT version anyone? Yes, they do exist.
.386 device driver anybody? Seem to remember them being around back when
Windows 3.1 was released too...
Ian Griffiths
: "Quantum" : A minimum amount of a physical quantity which can exist and by
: multiples of which changes in the quantity occur.
"Quantum leap" - a popular term derived from a misunderstanding of
Quantum theory, used to mean a massive leap; a Sci-fi series
Ian Griffiths
: Indeed, isn't the general idea to gradually aim everyone twoards
: their Cairo system (I do get confused with all these names, so I
: maybe wrong). In the meantime, is it Daytona that is Windows NT with
: a Windows 95 style interface?
Yes to the first bit, but I think that Daytona is just NT 3.5, which
looks a lot like normal Windows, only it doesn't crash.
Cairo is more like NT with a Win95 interface, only better.
Karl 'RiscMan' Davis
> In <D3Mqz...@festival.ed.ac.uk> a...@roe.ac.uk (Ale Terlevich) writes:
> >In article 2...@sallery.demon.co.uk, ga...@sallery.demon.co.uk (Gavin Sallery) writes:
> >>It's fast.
> > If you're saying a StrongARM is fast, how do you know? Have you seen one?
> >Do you have specs? Or are you just guessing.
> 4 times faster than the existing ARM610 and about the speed of the existing
> PowerPC.
Wooh! Yay! Excellent.. ARM strikes back! :)
--
Cheers!
+----------------------------+---------------------------------------------+
|.__ . . |---- You took the dog, you took the car, ----|
||__) o __ __ |\/| __ _ |--------- You left me on a monday; ----------|
|| \ (__ | `| | '__\ | ||----- The only thing you didn't take --------|
|| \ | ___)|__.| | (__) | ||---- Was my half-full bottle of Bundy. ------|
+-President of the Intel Removal Society, Regd off. ARMature II BBS Sydney-+
Tony Finch
Armstrong) wrote:
>
> yeah, but doesn't your average P90 instruction do more than a single
> alpha instruction? I was once told that alpha instructions were very
> simple beasts, but that the alpha executed them at silly speeds...anyone
> else with greater contact with alpha code than me willing to comment?
These days CISC's high-functionality-per-instruction isn't an advantage
because it ties up many parts of the processor on one instruction. It's
easier to analyse simpler instructions for clashes so that you can get
more parallelism from them. The Alpha instruction set is designed to help
this aim.
Tony.
--
If you don't know where I am by now, you've not been paying attention.
Ernest Ong
: In the press releases it was said that the StrongARM would be about 3 to
: 4 times faster than the existing ARM610 and about the speed of the existing
: PowerPC.
Did the press get any more specific about which existing PowerPC?
--
name: Ernie Ong
addr: er...@tartarus.uwa.edu.au
mesg: The Acorn Risc PC.
ARMed for the future.
Timothy Kimber
In article <3he48a$4...@oak70.doc.ic.ac.uk>, cp...@doc.ic.ac.uk
>
>If think a good description of what DEC are going to consider when
>designing various chips can be boiled down to:
>StrongARM family
>How fast can we make it, keeping it compatible (with other ARM's)
>having power consumption less than 1W, price less than 100 USD in bulk.
>(figures just my guesses)
Which begs the questions :
A) Why does the new chip need to be compatible with existing ARM chips ?
B) Why does DEC want a low power consumption chip ?
--
___
|im (cee...@cee.hw.ac.uk)
A Citizen of the Outside
Tony Finch
i...@intrac.demon.co.uk (Ian Griffiths) wrote:
>
> Cairo is more like NT with a Win95 interface, only better.
But it's still bloated. An operating system should not be like a blue
whale on a tricycle.
K.J. Bracey
>Thilo Manske (Thilo....@arbi.informatik.uni-oldenburg.de) wrote:
>> But do you really believe, you can go to your local dealer in summer 1996
>> and buy a StrongARM-Card for your RiscPC? I don't think so, look what
>> happens to the PC-Card, ARM700 or ARM800 cards? You should be able to buy
>> them all today, but in Germany not even the PC-Card has been released to
>> trade.
>> I said we will get a working MACHINE in five years, not a CPU.
>> There's a difference, you know.
>> --
>I know the feeling!, but if the Strong ARM is compatible with the
>ARM7/8 then presumably it wont need a ASIC or much extra logic to
>function with the RPC. Hopefully it will just be a matter of building
>the board, plugging a 100Mhz+ clock module on it, adding a few gates
>here and there and plugging it in.
> Anyway i'm sure i read somewhere Acorn werent going to support the
>RPC after a few years and would move on to a multiprocessor machine.
>If they did this then surely they'd go for the most cost-effective
>processor, which presumably should be the ARM800 then.
Erm, what do you think the RPC is then, apart from being a (1st generation)
multiprocessor machine? You can currently plug up to six processors into it.
Kevin
=====
Ollie Cornes
> In article <1995Feb11....@intrac.demon.co.uk>,
> i...@intrac.demon.co.uk (Ian Griffiths) wrote:
> >
> > Cairo is more like NT with a Win95 interface, only better.
>
> But it's still bloated. An operating system should not be like a blue
> whale on a tricycle.
'Bloated' implies that NT is bigger than it should be. For what it does,
I don't think you will get away with saying that. This is after all a
heavy duty networked workstation OS we're talking about here not an
OS for cheap machines.
Cheers,
Ollie.
--
\ /\ Computer Concepts Ltd Xara Ltd
( ) ol...@cconcepts.co.uk ol...@xara.co.uk
.( o ). All opinions are mine
K.J. Bracey
>In article <3he48a$4...@oak70.doc.ic.ac.uk>, cp...@doc.ic.ac.uk
>(Christopher Paul Rijk) writes:
>>
>>If think a good description of what DEC are going to consider when
>>designing various chips can be boiled down to:
>>StrongARM family
>>How fast can we make it, keeping it compatible (with other ARM's)
>>having power consumption less than 1W, price less than 100 USD in bulk.
>>(figures just my guesses)
>Which begs the questions :
>A) Why does the new chip need to be compatible with existing ARM chips ?
So you can plug it into your Risc PC, perhaps?
>B) Why does DEC want a low power consumption chip ?
Just trying to broaden their product range.
Kevin
=====
Keith Marlow
OB>Using a smaller die size boosts the yield though. The reason for this
OB>is
OB>that each defect on the wafer will probably make the chip it's in
OB>useless.
OB>So if the chips are smaller, you get more useful chips per wafer, and
OB>about
OB>the same number of duds (maybe a few more as it's less likely you'll
OB>get
OB>multiple defects per chip). So the %age yield is better.
Hmm, but isn't it also the case that a defect that was too small to cause a
problem at a larger size could cause a problem at a smaller size? Wouldn't the
defects per slice have to go down to get a better yield?
OB>Of course it's swings and roundabouts, like most things.
True, but in this case very *small* swings and roundabouts :-)
===========================================================================
| Keith Marlow | Efficient, | Supreme Software Systems Ltd. |
| sy...@supreme.demon.co.uk | Compact & | 21 Courtenay Close |
| Voice/Fax 01603 745077 | Fast.. | Norwich Norfolk |
| Data 01603 744231 | Windows??? | NR5 9LB |
===========================================================================
Wayne Bagguley
In article <fanf2-14029...@reshep.trin.cam.ac.uk>, fa...@cam.ac.uk (Tony
Finch) writes:
|> In article <1995Feb11....@intrac.demon.co.uk>,
|> i...@intrac.demon.co.uk (Ian Griffiths) wrote:
|> >
|> > Cairo is more like NT with a Win95 interface, only better.
|>
|> But it's still bloated. An operating system should not be like a blue
|> whale on a tricycle.
I suppose that depends on the size of the tricycle. ;-)
|>
|> Tony.
|> --
|> If you don't know where I am by now, you've not been paying attention.
Wayne.
Jasper the Apricot
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
X-Sender: lsuqf@lily
In-Reply-To: <kjb1003.2...@phx.cam.ac.uk>
On Mon, 13 Feb 1995, K.J. Bracey wrote:
> Date: Mon, 13 FEB 1995 17:55:52
> From: K.J. Bracey <kjb...@phx.cam.ac.uk>
> Newgroups: comp.sys.acorn
> Subject: Re: ARM and DEC developing StrongARM
>
> In article <3ho5ci$s...@yama.mcc.ac.uk> mbcaprt@PROBLEM_WITH_INEWS_GATEWAY_FILE (Robert Templeman) writes:
>
> >Thilo Manske (Thilo....@arbi.informatik.uni-oldenburg.de) wrote:
>
> > Anyway i'm sure i read somewhere Acorn werent going to support the
> >RPC after a few years and would move on to a multiprocessor machine.
Oh great! Acorn have only just dumped the vast majority of their users by
producing the damn RPC - and now by the time I get anywhere near the
amount of money needed to get even the base model, Acorn are going to
ditch this too?
*Please* tell me that I've misunderstood!
At least one advantage of having a PC is that you can be pretty sure that
you can upgrade your computer easily and without any fuss (I'm referring
to purchasing rather than installing here!) - and still be 'with it' as
far as your hardware is concerned...