700 Upgrade cards
Chris Cox
At the launch of the RiscPC we made a commitment to rolling program of
processor upgrades. To prepare for the release of the 700 upgrade cards
during the second quarter (ish) I want to make sure that we get the right
number and mixture. So I am looking for some indication of how many existing
users feel that they will upgrade to a new card, and whether it will be to
the 710 integer only card, or the higher cost 700 card with the FPA
(Floating Point Accelerator)
Details about the different processors were included in the original RiscPC
datasheets, and can be obtained from the ftp site in
pub/documents/products/RiscPC/technotes where they are available either
plain or sparkived.
I look forward to reading your comments.
Chris Cox
Enthusiasts Sales and Marketing Manager
Frank Jukes
cc...@acorn.co.uk (Chris Cox) wrote:
> Please could you help us to get it right!
>
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
Three questions.
I seem to remember that the difference in cost was to be 75ukp for the
'integer only' card, is that still the case?
Does anyone know what the cost of the FPA card might be?
How many months are there in an 'ish'?
Frank
--
Frank Jukes
e-mail, fr...@fjukes.demon.co.uk
... Dew knot trussed yore spell checquer two fined awl mistakes.
Sveinung W. Tengelsen
: Please could you help us to get it right!
: At the launch of the RiscPC we made a commitment to rolling program of
: processor upgrades. To prepare for the release of the 700 upgrade cards
: during the second quarter (ish) I want to make sure that we get the right
: number and mixture. So I am looking for some indication of how many existing
: users feel that they will upgrade to a new card, and whether it will be to
: the 710 integer only card, or the higher cost 700 card with the FPA
: (Floating Point Accelerator)
: datasheets, and can be obtained from the ftp site in
: pub/documents/products/RiscPC/technotes where they are available either
: plain or sparkived.
: I look forward to reading your comments.
: Chris Cox
I'd go for the one with FPA. By the way, I have three questions for you
(Chris Cox) in this regard;
1) When is it coming? A date, please.
2) What will it cost? Price on the original RPC datasheet 1 is L125.
Is this still correct, the L's inflation taken into account?
3) What is it's speed? Refering to the same datasheet, it's 'round 45 -
80 ARM MIPS?
I hope you can answer these questions here.
Bye, Pix. (.sig absent.)
s.p.beck
: How many months are there in an 'ish'?
: Frank
: --
: Frank Jukes
: e-mail, fr...@fjukes.demon.co.uk
: ... Dew knot trussed yore spell checquer two fined awl mistakes.
Well judging by the PCcard about 8 months in an ish......
--
Any views expressed here are entirly my own (honest!!), and do
not reflect the views of anyone else at all. Even then, I am
probably wrong...but don't quote me on that.......
Dr R.J. Gayton
: Please could you help us to get it right!
: At the launch of the RiscPC we made a commitment to rolling program of
: processor upgrades.
(Stuff deleted....)
You also made a commitment to PC cards in October '95
(Mind you, as I play with my newly-arrived 486 card :-) :-) I must say that
the wait was worthwhile. For *real-world* users (as against people who
just read spec-sheets, but never actually do anything with the kit) the
card is excellent.)
I think the most important thing for Acorn to bear in mind as ARM7xx time
draws near is that it is immensely harmful to their credibility to offer
what they cannot deliver. I know the PC-card fiasco was not altogether
of their making, but the way it was handled was a public-relations
disaster.
Rabin Ezra
core rather than the standard? If you're using an ints only machine, fixed
point is crippled by the standard multiply.
Rabin Ezra
[In a personal capacity]
--
Rabin Ezra E-mail: ra...@acm.org
Phone: +44 1483 448813
This e-mail may express different views to Fax: +44 1483 448811
those of Criterion Software Ltd. Home: +44 1483 202527
"Unix _IS_ user friendly... It's just selective about who its friends are."
Donald McCarthy
I have an RPC600 and need the floating point performance for my simulator.
I would get the FP unit if it is affordable, but would also be interested
in even faster FP support
Cheers, Paddy.
NOTE: My thoughts, Fujitsus' computers.
sjw
> Please could you help us to get it right!
> [snip]
>
> Details about the different processors were included in the original RiscPC
> datasheets, and can be obtained from the ftp site in
> pub/documents/products/RiscPC/technotes where they are available either
> plain or sparkived.
>
Weren't the prices guaranteed for 5 years of which we have already had 1
with no upgrade available!
Malcolm Lithgow
: cc...@acorn.co.uk (Chris Cox) writes:
: > At the launch of the RiscPC we made a commitment to rolling program of
: > processor upgrades. To prepare for the release of the 700 upgrade cards
: > during the second quarter (ish) I want to make sure that we get the right
: > number and mixture. So I am looking for some indication of how many existing
: > users feel that they will upgrade to a new card, and whether it will be to
: > the 710 integer only card, or the higher cost 700 card with the FPA
: > (Floating Point Accelerator)
: The 700 FPA, no question.
: Every so often I need to do floating point work and the software emulation
: makes that a little slow. And I want more processor grunt. :)
Me too. I'm not sure how much I'd use it (most of the work I do on the
Risc PC is DTP), but I couldn't see much point in upgrading to just a
710.
-Malcolm. lit...@uslp.usl.com
Philip R. Banks
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
The 700 FPA, no question.
Every so often I need to do floating point work and the software emulation
makes that a little slow. And I want more processor grunt. :)
Philip
--
Philip R. Banks Syntax: mail < ban...@kosmos.wcc.govt.nz > @@@@@/|
Take a look around you at the world we've come to know, @@@@/#|
Does it seem much more than a crazy circus show? @@@/##|
Maybe from the madness something beautiful will grow.... @@/---|
---The Artilleryman from Jeff Wayne's "War of the Worlds" @/ |
Steve Pringle
> Please could you help us to get it right!
>
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
>
> I look forward to reading your comments.
>
> Chris Cox
> Enthusiasts Sales and Marketing Manager
>
I would definitely buy the 700 FPA card. In fact, I wonder if standardisation on the 700 FPA instead of producing both 710 and 700 based cards would lead to a lower cost 700 card (volume of sales etc.)?
--
Steve
-----
Steve Pringle
spri...@latrigg.demon.co.uk
Simon Burrows
> So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
Clocked at the same speed?
How much difference in price..?
--Simon
Andrew Matthews
> cc...@acorn.co.uk (Chris Cox) writes:
>
> > At the launch of the RiscPC we made a commitment to rolling program of
> > processor upgrades. To prepare for the release of the 700 upgrade cards
> > during the second quarter (ish) I want to make sure that we get the right
> > number and mixture. So I am looking for some indication of how many existing
> > users feel that they will upgrade to a new card, and whether it will be to
> > the 710 integer only card, or the higher cost 700 card with the FPA
> > (Floating Point Accelerator)
>
> The 700 FPA, no question.
I don't own a RPC (yet). However as some-one else said,, the prices of the
upgrades were meant to be held for a few years for exisiting owners. However
one year has gone by and no upgrade, and no indication of one.
I will wait to replace my A5000 with a 700 (I want the FPA as well), so I
won't be buying until it appears (No, I am not going to hold my breath).
--
Andy
-----------------------------------------------------------
Some people can tell what time it is by looking at the Sun,
but I have never been able to make out the numbers.
-----------------------------------------------------------
... Plasma is another matter.
Richard West
|> Please could you help us to get it right!
|>
|> processor upgrades. To prepare for the release of the 700 upgrade cards
|> during the second quarter (ish) I want to make sure that we get the right
|> number and mixture. So I am looking for some indication of how many existing
|> users feel that they will upgrade to a new card, and whether it will be to
|> the 710 integer only card, or the higher cost 700 card with the FPA
|> (Floating Point Accelerator)
|>
[This man probably wants replies by e-mail, but I'm going to air my views in
public].
The answer is "depends". Depends on what? Depends on:
1. How much (money)? The quoted figure was 125UKP+VAT to upgrade to a >40MHz
ARM700 (which I presume means 55MHz, and I presume means integer only). How
much extra performance do I get? A factor two and I won't bother, a factor
five and I probably would.
2. Will it be the 'DM' variant of the 700/710? If not, why not?
3. Would the ARM700 card be available with an empty FPA socket?
4. Would such a card take an FPA10, and if not, why not? (Clock speed?) I've
already bought one FPA, I don't really want to have to buy another one.
5. [When] Will the ARM800 cards be available? If you say 1996 Q1 then there's
no point wasting my cash on a 700 card...
Yours, trying to extract more information from Acorn/blood from a stone...
Ian Rawlings
cc...@acorn.co.uk (Chris Cox) wrote:
>Please could you help us to get it right!
>number and mixture. So I am looking for some indication of how many existing
>users feel that they will upgrade to a new card, and whether it will be to
>the 710 integer only card, or the higher cost 700 card with the FPA
>(Floating Point Accelerator)
Well, I'm not an existing user as I'm not buying an RPC until it offers
something above my 8Mb A5000, ie a faster processor and decent FP. When
will machines be released with this as standard? I'd like the DM core
with an FP preferrably :-)
Ian
--
Opinions? I am a Fortean!
Ian Rawlings, i...@daffodif.demon.co.uk - England.
C.N.Good
: At the launch of the RiscPC we made a commitment to rolling program of
: processor upgrades. To prepare for the release of the 700 upgrade cards
: during the second quarter (ish) I want to make sure that we get the right
: users feel that they will upgrade to a new card, and whether it will be to
: the 710 integer only card, or the higher cost 700 card with the FPA
: (Floating Point Accelerator)
Its worth bearing in mind that the readers of this group are distinctly
different from the bulk of acorns market and as such the proportion of
c.s.a.* readers wanting arm700 +FPA is going t be much greater than the
rest of the market.
Chris
--
+-----------------------------------------------+------------------------+
| Chris Good (Undergraduate - Computer Science) | Intel => Fool |
| University Of Kent at Canterbury, UK | Inside => Outside |
| Take no notice, I speak for no-one | ARM Powered |
+-----------------------------------------------+------------------------+
D I Simms
|> Philip R. Banks (ban...@khantazi.central.co.nz) wrote:
|> : cc...@acorn.co.uk (Chris Cox) writes:
|>
|> : > processor upgrades. To prepare for the release of the 700 upgrade cards
|> : > during the second quarter (ish) I want to make sure that we get the right
|> : > number and mixture. So I am looking for some indication of how many existing
|> : > users feel that they will upgrade to a new card, and whether it will be to
|> : > the 710 integer only card, or the higher cost 700 card with the FPA
|> : > (Floating Point Accelerator)
|>
|>
|> : Every so often I need to do floating point work and the software emulation
|> : makes that a little slow. And I want more processor grunt. :)
|>
|> Risc PC is DTP), but I couldn't see much point in upgrading to just a
|> 710.
|>
You must be joking. The problem with the 700 is that it cannot be clocked as
highly as the 710. I would prefer a mega fast ARM710 - perhaps a 55Mhz one
like someone had at the last show :-). The highest clocked ARM700 that I have
heard of is 40Mhz.
After all, which applications will benefit from the floating point. Impression
and Artworks certainly wont.
Dan Simms.
Roger Woodhouse
lit...@summit.novell.com (Malcolm Lithgow) wrote:
> Philip R. Banks (ban...@khantazi.central.co.nz) wrote:
> : cc...@acorn.co.uk (Chris Cox) writes:
>
> : > At the launch of the RiscPC we made a commitment to rolling program of
> : > processor upgrades. To prepare for the release of the 700 upgrade cards
> : > during the second quarter (ish) I want to make sure that we get the right
> : > number and mixture. So I am looking for some indication of how many existing
> : > users feel that they will upgrade to a new card, and whether it will be to
> : > the 710 integer only card, or the higher cost 700 card with the FPA
> : > (Floating Point Accelerator)
>
> : The 700 FPA, no question.
>
> : Every so often I need to do floating point work and the software emulation
> : makes that a little slow. And I want more processor grunt. :)
>
> Me too. I'm not sure how much I'd use it (most of the work I do on the
> Risc PC is DTP), but I couldn't see much point in upgrading to just a
> 710.
>
Could some techy :-) person tell me what the advantages are in a FPA?
(ie. what type of work does it improve? ie. dtp etc.?)
Thanks
Roger
--
... If I want your opinion I'll beat it from you
Stephen Parkin
> Please could you help us to get it right!
>
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
>
Personally, I don't think I'd bother with the 710.
--
Stephen
Ernest Ong
: |>Please could you help us to get it right!
: |>
: |>At the launch of the RiscPC we made a commitment to rolling program of
: |>processor upgrades. To prepare for the release of the 700 upgrade cards
: |>during the second quarter (ish) I want to make sure that we get the right
: |>number and mixture. So I am looking for some indication of how many existing
: |>users feel that they will upgrade to a new card, and whether it will be to
: |>the 710 integer only card, or the higher cost 700 card with the FPA
: |>(Floating Point Accelerator)
In fact, the _only_ reason why I, and a few others, are holding back
from buying a Risc PC at this point, is precisely because of the lack of
the ARM 7X0 card. Sure we can upgrade it, but we figured it'd be cheaper
to simply buy a Risc PC 7X0 than a Risc PC 600 _then_ upgrade (even if
the upgrade is cheap). Some of us are tight on the $$$... :-(
I obviously can't speak on behalf of others, but I see the ARM 7X0 card
as a vital addition which must arrive in the very near future. As
regards whether the 710 or 700 card - how about a 700 card withOUT an
FPA but a cheap (?) upgrade option to the FPA? Some people (like me)
might not realise they might need an FPA (perhaps due to a change to the
applications they use) when they get the machine, and it's nice to have
the option open to get an FPA when they need it. So the answer, on my
part, would be: a 700 card with OUT an FPA, but a cheap upgrade to the
FPA.
cheers,
ernie.
--
name: Ernie Ong
addr: er...@tartarus.uwa.edu.au
mesg: The Acorn Risc PC.
ARMed for the future.
Richard Peter Samuels
[YANK, SNIP]
> Chris Cox
> Enthusiasts Sales and Marketing Manager
As the 700 is (I believe) only a _bit_ quicker on intege maths than
the 610, It would make sense to upgrade to 700+FPA rather than 710 as
it seems the only real advanage is the ability to have floating point in
hardware.
Trev
experiment with. I was using it on my RiscPC with (presumably) an ARM610.
For the simple stuff, it was fine. But anything complex took ages.
Hence, I wiped it until I've got a more powerful computer.
Now, I don't know much about raytracing, though I am aware that it's not
unusual for pictures to take hours to render.
The docs say that POVRay supports floating point maths, so would an ARM710
with FPA provide any significant speed increases? And how would it
compare to a 486/Pentium doing similar tasks?
Do Acorn have any speed tests for actually running applications?
Thanks,
Trev
--
_____________________________________________________________
`\|||/'o O / Trevor K. Johnson, email: t.k.j...@bradford.ac.uk \
(o o) | Civ. Eng. Student, CD ..WWW page pending.. |
ooO_(_)_Ooo__\_collector_and_Acorn_Anorak_Extraordinaire___________________/
R.E.O'Brien
> Please could you help us to get it right!
>
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
>
- SNIP -
> Chris Cox
> Enthusiasts Sales and Marketing Manager
>
I don't yet have a RiscPC. The only remaining reason for this
is that I don't want to give up the floating point performance I
have with my A5000 + FPA!. As soon as a RiscPC with a 700 card plus
FPA is available, I will order one!
--
Rob O'Brien
Matthew Wilcox
: At the launch of the RiscPC we made a commitment to rolling program of
: processor upgrades. To prepare for the release of the 700 upgrade cards
: during the second quarter (ish) I want to make sure that we get the right
: number and mixture. So I am looking for some indication of how many existing
: users feel that they will upgrade to a new card, and whether it will be to
: the 710 integer only card, or the higher cost 700 card with the FPA
: (Floating Point Accelerator)
Presumably, the integer only unit will be clocked faster as well. Will
there be some sort of speed selection? ie test all the processors at (say)
60 MHz then those that fail at 50 MHz then the others at 40 MHz. This would
make a lot of people happier, I think. Is there any 'best guess' figure as
to how fast the Arm 7's will be clocked, by the way? I know the figures
that Acorn User published but after that awful article they published by
Chris Mellor, I'm not sure how far I can trust anything they say.
--
+=======================================================================+
| "We're bingo numbers and our names | If the University cared about /|
| are obsolete; Why do I feel bitter | my opinions, they wouldn't / |
| when I should be feeling sweet?" | be teaching me Ada. /--|
| Hello, turn your radio on - S.Sister| M. Wilcox mrw...@york.ac.uk / |
+=======================================================================+
Frank Jukes
cc...@acorn.co.uk (Chris Cox) wrote:
> Please could you help us to get it right!
[snip]
> I look forward to reading your comments.
We've asked you quite a few questions in the news groups, could we have
replies please?
Copied to C Cox by email.
Frank
--
Frank Jukes
e-mail, fr...@fjukes.demon.co.uk
... It's dangerous to be right when the government is wrong. -Voltaire-
Steve Holroyd
fr...@fjukes.demon.co.uk (Frank Jukes) wrote:
I think it is a tad unreasonable to expect a reply over the Easter
weekend! I suspect this is not what you mean however :-)
--
Steve Holroyd
Ollie Cornes
> Could some techy :-) person tell me what the advantages are in a FPA?
> (ie. what type of work does it improve? ie. dtp etc.?)
Not far off no improvement at all at the moment. There's
very few apps that would go any faster at all. Though
if machines start selling with 'em in, people will take
advantage of it, rather than messing about with the
delights of fixed point integer maths (math to those of you
reading in American).
The main advantage I can see is that it would be feaible to
run a ray-tracer on an arc! Using a machine without it
for this is laughable.
Regards,
Ollie.
--
\ /\ Computer Concepts Ltd Xara Ltd
( ) ol...@cconcepts.co.uk ol...@xara.co.uk
.( o ). All opinions are mine
Matthew Wild
> Please could you help us to get it right!
>
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
>
> I look forward to reading your comments.
>
> Chris Cox
> Enthusiasts Sales and Marketing Manager
>
> I would definitely buy the 700 FPA card. In fact, I wonder if standardisation
> on the 700 FPA instead of producing both 710 and 700 based cards would lead to
> a lower cost 700 card (volume of sales etc.)?
>
> --
> Steve
> -----
> Steve Pringle
> spri...@latrigg.demon.co.uk
I agree. I would have thought that with what is hopefully a minimal cost difference
between the 710 and 700, it would make more sense to just produce 700's for the
RISC-PC market, and let the 710's be used in the embedded markets for which they have
some use.
Matthew
--
Matthew Wild
M.W...@rl.ac.uk
World Data Centre C1 - Solar-Terrestrial Physics
Rutherford Appleton Laboratory, Chilton, Didcot, Oxon, OX11 0QX
Philip R. Banks
> As the 700 is (I believe) only a _bit_ quicker on intege maths than
> the 610, It would make sense to upgrade to 700+FPA rather than 710 as
> it seems the only real advanage is the ability to have floating point in
> hardware.
The 700 has, if I remember correctly, a clock halver for the co-processor
interface that allows a 50MHz 700 to be paired with a 25MHz FPA. The extra
clock speed and slightly better integer performance coupled with an FPA should
be very nice.
Philip
--
Philip R. Banks Syntax: mail < ban...@kosmos.wcc.govt.nz > @@@@@/|
The History of the Universe in three Acts. @@@@/#|
Act I) *Bang!* Act II) *hiSSSsssss....* Act III) *Crunch!* @@@/##|
@@/---|
--from "The State of the Art" by Iain M. Banks. @/ |
Olly Betts
> Its worth bearing in mind that the readers of this group are distinctly
>different from the bulk of acorns market and as such the proportion of
>c.s.a.* readers wanting arm700 +FPA is going t be much greater than the
>rest of the market.
The folks round here are mostly techies, and the sort of people to buy
new-fangled whizz-bang gadgets (or maybe it's just me). So we're probably a
fairer sample of the sort of people who would rush out and buy a faster
processor card. We're the early demand, which is what Chris Cox was trying
to judge (as I read it anyway).
If I were to upgrade, I'd probably buy an ARM710 if the ARM700 was clocked
appreciably lower, and hence slower in general use. Whether I upgrade at
all depends on how much faster it is than an ARM610 and what colour my bank
balance is at the time.
Olly
--
Reality is for people who can't cope with reality.
Ian Trimble
> The ARM700 is noticeably more expensive to produce than the ARM710
> simply because it has more pins, which pushes up the packaging cost.
How much more? UKP30 v 20?
I thought the whole point was that ARM chips are cheap to produce
compared to beermat size Pentiums etc.
To stick anything but the fastest possible ARM chip in the next
iteration of the RiscPC would be the utmost folly.
__ ____
/ _ _ / _ ' _ _ / / _
___/__(_|_| )___/_/ _(_| ) )_/_)_(_(/___________________________________
Dr Ian Trimble email: tr...@sherwood.demon.co.uk
Sherwood Health Centre
Elmswood Gardens Tel: +44 115 962 4516
Nottingham NG5 4AD Fax: +44 115 985 7899
________________________________________________________________________
Ian Trimble
> The 700 has, if I remember correctly, a clock halver for the co-processor
> interface that allows a 50MHz 700 to be paired with a 25MHz FPA. The extra
> clock speed and slightly better integer performance coupled with an FPA
> should be very nice.
At last year's Harrogate show I was told that my 25MHz FPA10 would not
work with an ARM 700 due to incompatible clock speeds. The ARM 700 would
instead be paired with the 33MHz FPA11.
Now this could imply an ARM 700 clock speed of either 66MHz or, more
likely, 50MHz. Any bets?
I would like to think that the long gestation of the ARM 700 has been
due to ARM's efforts to produce a reliable 66MHz chip, but I'm probably
wrong :-(
Ian Trimble
> I will wait to replace my A5000 with a 700 (I want the FPA as well), so I
> won't be buying until it appears (No, I am not going to hold my breath).
Seconded.
IMHO the RiscPC 700 series should include FPA as standard. A rapidly
swelling number of FPA owners might persuade software developers to
upgrade their products - particularly spreadsheets (the Acorn does feel
rather sluggish in the spreadsheet department).
Simon Stevens
>Please could you help us to get it right!
>
>At the launch of the RiscPC we made a commitment to rolling program of
>processor upgrades. To prepare for the release of the 700 upgrade cards
>during the second quarter (ish) I want to make sure that we get the right
>number and mixture.
>
>
Erm, this may be a petty comment, but I'm STILL waiting for my 486 PC card to
arrive - I only bought my RiscPC last July, and it was promised that it would
arrive by October 94 !! (Beebug say they are releasing them "As soon as Acorn
send them to them"). Once the PC card does arrive, then I might think about
buying another upgrade ( but if the PC card fiasco is anything to go by I might
just not bother... Even Intel can get hardware released with only a 6 month
delay...!!)
Reality Is An Illusion Caused By A Lack Of Reality ... !"
==================================================================
Simon Stevens Email : stev...@sisko.sci.port.ac.uk
Computer Support Technician or stev...@sci1.sci.port.ac.uk
Science Faculty
University Of Portsmouth
D I Simms
My thoughts exactly. The only way I would buy an ARM700 is if it was clocked
highly with the FPA clocked at half its speed. Say a 50/60Mhz ARM700 with a
25/30Mhz FPA.
I mean RiscOS does not even use the FPA so the OS would not be any faster. It
all seems to hinge around the people who use cad. I would certainly prefer
a highly clocked integer chip which benefits RiscOS.
Dan Simms.
PS Where I use the term RiscOS above I mean the actual operating system and not
the applications provided with it.
Richard West
(Ian Trimble) writes:
|> In message <A28QU...@khantazi.central.co.nz> Philip R. Banks wrote:
|>
|> > The 700 has, if I remember correctly, a clock halver for the co-processor
|> > interface that allows a 50MHz 700 to be paired with a 25MHz FPA. The extra
|> > clock speed and slightly better integer performance coupled with an FPA
|> > should be very nice.
|>
|> At last year's Harrogate show I was told that my 25MHz FPA10 would not
|> work with an ARM 700 due to incompatible clock speeds. The ARM 700 would
|> instead be paired with the 33MHz FPA11.
|>
|> Now this could imply an ARM 700 clock speed of either 66MHz or, more
|> likely, 50MHz. Any bets?
|>
|> I would like to think that the long gestation of the ARM 700 has been
|> due to ARM's efforts to produce a reliable 66MHz chip, but I'm probably
|> wrong :-(
|>
Funny how we're not told this information, isn't it?
If it were 50MHz, and clock halving is available, then surely an FPA10 would
work with it? (I believe I recall correctly that the FPA10 runs at 25MHz, or am
I wrong???)
Given the FPA10 costs ~115 quid, presumably the mythical FPA11 will cost more.
Add that to the ~150 that the 700 board would cost, you must be looking at ~300
pounds.
How about a 50Mhz ARM700 board with an empty socket for the FPA I've already
got??? Well, I can hope...
Chris Cox
that they would upgrade to the 700 cards, and which one. There have been a
number of questions raised, and I have put together the following document
to try and pull together the answers:
Q. Weren't the prices guaranteed for 5 years of which we have already had 1
with no upgrade available!
A. The upgrade prices are guaranteed (subject to RPI) there was no time
limit specified. The upgrade from the standard 610 to a basic 40MHz 710
card was quoted at #100
Q. What the additional cost of the FPA card?
A. The FPA prices have not yet been set, but should be around #50 more than
the 710 card price
Q. How many months are there in an 'ish'?
A. Should be zero but could be more if we hit any unforseen problems
Q. Will it be the 'DM' variant of the 700/710? If not, why not?
A. There not be a DM version of the 700/710 card as there is no need for the
extra cost an complexity of the D (Debug) version in a product, and there
is no ARM7 M version available
Q. Would the ARM700 card be available with an empty FPA socket?
A. No, there will be two cards one with the 700 and FPA11 and one with the
710 which cannot use the FPA. There will not be a version with a 700
without an FPA11 fitted
Q. When will the ARM800 cards be available?
A. The process for developing a card is that the chip is developed and then
the a prototype chip is made, this is then used to develop a prototype
board, and any modifications to the chip are then incorporated before
production parts are available, these can then be built into production
boards. At the moment the 800 chip is in the design stage, so any
estimate of a completion date for the board would be very likely to
change, as it is very hard to predict what modifications will be needed
Q. Are both cards clocked at the same speed?
A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
be clocked at 33MHz
Q. What are the benefits of using an FPA processor
A. Some applications such as Raytracing use a lot of floating point
arithemtic, these applications will be greatly speed up if they can use a
hardware floating point unit
Ale Terlevich
>In message <A28QU...@khantazi.central.co.nz> Philip R. Banks wrote:
>
>> The 700 has, if I remember correctly, a clock halver for the co-processor
>> interface that allows a 50MHz 700 to be paired with a 25MHz FPA. The extra
>> clock speed and slightly better integer performance coupled with an FPA
>> should be very nice.
>
>At last year's Harrogate show I was told that my 25MHz FPA10 would not
>work with an ARM 700 due to incompatible clock speeds. The ARM 700 would
>instead be paired with the 33MHz FPA11.
>
>Now this could imply an ARM 700 clock speed of either 66MHz or, more
>likely, 50MHz. Any bets?
HUH?
Ok. The ARM700 can use a copro clocked at the same speed, or at half
the speed of itself (If I remember this thread from a few months back
correctly) so how do you get a 50MHz ARM700 from a 33MHz FPA10? I'd
have thought that if the FPA11 was going to be clocked at 33 MHz then
the ARM700 would be 33 or 66MHz, and if the ARM700 was going to be 50MHz, then
we could use an FPA10.
If the ARM700+FPA runs at 33MHz, I will be *******MOST******* disappointed.
>
>I would like to think that the long gestation of the ARM 700 has been
>due to ARM's efforts to produce a reliable 66MHz chip, but I'm probably
>wrong :-(
The official explanation was that Acorn were pushed out of the chip
manufacturing queue by the big boys (due at least in part to the Cobe
earthquake and pentium bug).
Ale.
John Morley
> The ARM700 is noticeably more expensive to produce than the ARM710
> simply because it has more pins, which pushes up the packaging cost.
Isn't it about time you integrated the FPA onto the same die as the ARM,
then. Compared with other processor designs ARM floating point performance
is embarrassing.
--
John.
Ian Griffiths
: Yesterday I dabbled with POVRay, as raytracing is an area I'd like to
: experiment with. I was using it on my RiscPC with (presumably) an ARM610.
: ...
: Now, I don't know much about raytracing, though I am aware that it's not
: unusual for pictures to take hours to render.
: The docs say that POVRay supports floating point maths, so would an ARM710
: with FPA provide any significant speed increases? And how would it
: compare to a 486/Pentium doing similar tasks?
For a 486, it depends on what kind. A 486 SX is probably going to be
a similar speed to an ARM610, or a bit slower, depending on the clock rate
of the 486. A 486 DX however is going to leave it standing, because it
has floating point hardware, and POVRay spends most of its time doing
floating point calculations.
It took about half an hour to do that chess demo picture at 640x480 on
my 486 DX2-66.
A Pentium is likely to me much faster since the floating point unit is
more efficient. However it's also quite likely to give you a different
result unless you have a recent chip due to bugs in the FP unit.. :-)
--
Ian Griffiths - writing from but not for Madge Networks Limited
Jon Rijk
Matthew Wild <wi...@wdcc1a.bnsc.rl.ac.uk> wrote:
>>I would definitely buy the 700 FPA card. In fact, I wonder if
>>standardisation on the 700 FPA instead of producing both 710 and 700 based
>>cards would lead to a lower cost 700 card (volume of sales etc.)?
>I agree. I would have thought that with what is hopefully a minimal cost
>difference between the 710 and 700, it would make more sense to just
>produce 700's for the RISC-PC market, and let the 710's be used in the
>embedded markets for which they have some use.
I'd go for the 700 FPA as when I want floating point I want high
performance.
Standardising on the 700 would hopefully make people who want to do a lot
of mathematical analysis actually consider the RPC as a viable system,
as currently I'm sure it's not given any thought, and I'll bet there are
plenty out there that want such systems.
--
Jon Rijk If nothing seems to be going wrong
you have overlooked something!
ja...@ecs.soton.ac.uk Murphy's 3rd Law
http://whirligig.ecs.soton.ac.uk/~jar92
Jon Rijk
D I Simms <csz...@scs.leeds.ac.uk> wrote:
>|> > Its worth bearing in mind that the readers of this group are distinctly
>|> >different from the bulk of acorns market and as such the proportion of
>|> The folks round here are mostly techies, and the sort of people to buy
>|> new-fangled whizz-bang gadgets (or maybe it's just me). So we're probably a
>|> fairer sample of the sort of people who would rush out and buy a faster
>|> processor card. We're the early demand, which is what Chris Cox was trying
>|> to judge (as I read it anyway).
... I definitely fit into this category.
[snip]
>My thoughts exactly. The only way I would buy an ARM700 is if it was clocked
>highly with the FPA clocked at half its speed. Say a 50/60Mhz ARM700 with a
>25/30Mhz FPA.
Yup, although I'd like to have floating point (and I mean >>4MFLOPS), I
would want what's going to make my machine as fast as possible, Acorns
machines are seriously lacking performance these days it's just the
applications/OS are written well so it don't seem so bad from a users point
of view.
Having an FPA at the moment won't be particularly useful but if it becomes
standard then things would be different as is said elsewhere.
Thorsten Seitz
>
>A. The FPA prices have not yet been set, but should be around #50 more than
> the 710 card price
>
That sounds good.
>A. There not be a DM version of the 700/710 card as there is no need for the
> extra cost an complexity of the D (Debug) version in a product, and there
> is no ARM7 M version available
How much extra cost would this be?
> boards. At the moment the 800 chip is in the design stage, so any
I would like to get more information about the ARM800 design, if possible.
>A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> be clocked at 33MHz
This is the point that worries me. Why is it not possible to clock the
ARM700 at 66MHz (or more) ???
Pentium/PowerPC are breaking the 100MHz barrier these days.
What is the state of StrongARM ?
Is there any information available ?
-Thorsten
Henry
> be clocked at 33MHz
and Ale Terlevich (a...@roe.ac.uk) wrote:
> If the ARM700+FPA runs at 33MHz, I will be *******MOST******* disappointed.
Pass the valium :(
We can spend an extra 50 quid, and buy a SLOWER machine - I think i've fallen
into a parallel universe.
Morgan,
--
I guess it was all a DREAM ... or an episode of HAWAII FIVE-O ...
Robert Templeman
> Please could you help us to get it right!
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
> Details about the different processors were included in the original RiscPC
> datasheets, and can be obtained from the ftp site in
> pub/documents/products/RiscPC/technotes where they are available either
> plain or sparkived.
> I look forward to reading your comments.
> Chris Cox
> Enthusiasts Sales and Marketing Manager
I will upgrade from my 410/1 when ever a RISC PC with an FPU becomes
available so add another FPA & ARM700 onto the production list 8-).
What might put me off buying one, though, is a low clock speed, at the moment
the RISC PC600 only offers better graphics performance than my 410,
which is ARM3'ed and over clocked.
A nice touch would be a small number of very fast ARM700's i know i
would pay alot extra for the speed.
Robert Templeman
Ian Trimble
> Q. Are both cards clocked at the same speed?
>
> A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> be clocked at 33MHz
According to the Acorn Processor Technology roadmap a 40MHz ARM 710
should be around 40% faster than a 30MHz ARM 610. A 33MHz ARM 700 should
be about 15% faster than an ARM 610.
The ARM 800 looks way off: Chris implies that a prototype chip has not
yet been made. We shall probably see a faster (66MHz) ARM 700 first
(possibly in the RiscPC 700?).
Could a technophile please enlighten me as to why achieving high clock
speeds for ARM chips is apparently so difficult? Is it because:
a) ARM don't have access to 0.5 micron fabrication?
b) transferring to a smaller die size is technically very difficult?
c) transferring to a smaller die size is very expensive?
d) chips running 4x faster than the bus would require faster memory
e.g. secondary cache on the processor card?
e) something else?
Dave Bass
[munch]
> It took about half an hour to do that chess demo picture at 640x480 on
> my 486 DX2-66.
> A Pentium is likely to me much faster since the floating point unit is
> more efficient. However it's also quite likely to give you a different
> result unless you have a recent chip due to bugs in the FP unit.. :-)
>
As I mentioned before, the P75 that I have did the job at 800x600 in under
nine minutes. The Pentium has a much improved FPU (the occasional bug
aside ;-).
(My A410/1 ARM3 + FPA10 @25MHz took 1 hr and some-odd minutes with the same
parameters)
You have to be a bit careful about making the comparison, though. POV has so
many options that can affect performance that saying 'that chess demo @
<state your resolution here>' is not sufficient.
Now, I gave up waiting for an RPC with ARM700 + FPA11 because of all the
delays, the cost and the prospect of only incremental performance improvements.
It seems I was right, as a 33MHz ARM700 + FPA11 seems unlikely to develop
significantly more Mega-flops than the old 25MHz combination.
It's a great pity as I am just getting back into the habit of using my Archie.
--
David Bass, ESTEC/YCV, Postbus 299, 2200 AG Noordwijk, The Netherlands
#include std_disclaimer
D I Simms
[snip]
|> Q. Are both cards clocked at the same speed?
|>
|> A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
|> be clocked at 33MHz
I seriously think Acorn have missed the general mood of the Acorn community.
Who really wants a 40Mhz ARM710.
If 40Mhz is the highest spec chip they can produce I think I will wait until
the ARM800/810 or the StrongARM is available. In fact I think I would rather
invest the extra UKP100 on a faster PC card.
Dan Simms. [who wants an ARM710 @ 55Mhz]
Douglas Pierce-Price
> I seriously think Acorn have missed the general mood of the Acorn community.
> Who really wants a 40Mhz ARM710.
>
> If 40Mhz is the highest spec chip they can produce I think I will wait until
> the ARM800/810 or the StrongARM is available.
Likewise --- for what appears to be a (relatively) small speed increase, even
with the slightly faster clocked ARM710, it hardly seems worth paying for the
upgrade. I will almost certainly wait for an ARM8xx (with some FP support) or
StrongARM based card. Buying something like the cards mentioned by Acorn
would be just a stop-gap.
--
Douglas Pierce-Price
Peter Burwood
> In message <3n2muh$2...@acorn.acorn.co.uk> Chris Cox wrote:
>
> > Q. Are both cards clocked at the same speed?
> >
> > A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> > be clocked at 33MHz
>
> According to the Acorn Processor Technology roadmap a 40MHz ARM 710
> should be around 40% faster than a 30MHz ARM 610. A 33MHz ARM 700 should
> be about 15% faster than an ARM 610.
I agree, those are the figures suggested by the graph in the technical
specification flyers that were distributed by Acorn. However, those are
for the dhrystone benchmark which is greatly effected by cache size.
What I, and dare I say everyone else here, want to know is what the
typical increase will be for processor intensive applications, such as a
compilation of a large system.
So, here is the challenge.
How long does the C compiler take to compile itself on a RPC fitted with
a 40Mhz 710 compared to a RPC fitted with a 30Mhz 610 ? Similarly for
the 33Mhz 700. (I'm not after absolute times, just ratios).
With the costs and specifications given I won't be bothering with the
700+FPA even though I was interested before. Role on the ARM 800.
> The ARM 800 looks way off: Chris implies that a prototype chip has not
> yet been made. We shall probably see a faster (66MHz) ARM 700 first
> (possibly in the RiscPC 700?).
If Acorn are going to put a 66Mhz ARM 700 in the RiscPC 700 then I want
it available as an upgrade now (assuming of course the RiscPC 700 is
going to appear shortly). If Acorn stuff us RiscPC 610 owners by not
allowing us to upgrade to similar spec machines then there are going to
a lot of pissed people.
regards,
Pete
-- Email: p...@arcangel.demon.co.uk
Ian Rawlings
Clive...@armltd.co.uk (you) wrote:
>In article <19950414....@daffodif.demon.co.uk> i...@daffodif.demon.co.uk writes:
>>Well, I'm not an existing user as I'm not buying an RPC until it offers
>>something above my 8Mb A5000, ie a faster processor and decent FP.
>
>Er... but the RiscPC already offers a lot over an 8Mb A5000. It's
>probably noticeably faster, because of VRAM, the write buffer, and
>16MHz main RAM instead of 12MHz, even though the processor is clocked
>no faster.
Err, perhaps, but if I sold my A5K, I'd get about £800-£900 tops
(probably optomistic, I haven't looked for ages!) and consequently only
be able to buy the base RPC!
Having said that, I still don't think the RPC is that hot.. I'm hanging
around in the hope that multiple processors will arrive one day.
--
Opinions? I am a Fortean!
Ian Rawlings, i...@daffodif.demon.co.uk - England.
Steve Holroyd
a...@roe.ac.uk (Ale Terlevich) wrote:
>
> >Q. Are both cards clocked at the same speed?
> >
> >A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> > be clocked at 33MHz
>
>
> Oh dear.
>
>
I may get shot down in flames here :-)
As I recall the situation, the clock speed of an ARM processor does not give
a good comparator for performance when matched against other processors. I
am sure that some of the responses to the news about the clock speeds of the
ARM 700/710 have been based on this fallacy.
The assumption that a 40MHz 710 will only perform marginally quicker than an
ARM3 (33MHz) based A5000 is not sustainable in my view.
I am impressed that POVray performance is now deemed to be an accurate
benchmark for comparisons to be made. Shame I don't use it :-)
Follow-ups and flames to comp.sys.acorn.hardware and comp.sys.acorn.misc -
why was the original mailing to comp.sys.acorn anyway? It is due to be
rmgrouped in another month or so. For some obscure reason, the "dead but not
yet buried" group is STILL the most popular. Have Acorn not got their new
groups sorted out yet?
--
Steve Holroyd
Garry
cc...@acorn.co.uk (Chris Cox) wrote:
[Snip]
> Q. Are both cards clocked at the same speed?
>
> A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> be clocked at 33MHz
Hmm... so for the average peasant (no ray tracing, no whizzy maths), it
will be better to buy a 710 (no fpu). Odd.
Well, I will buy one in an ish or two :-)
--
Garry
G4FRO @ GB7IMB.#41.GBR.EU
ga...@goldhill.demon.co.uk
Dr M.R. Clark
>Chris Cox (cc...@acorn.co.uk) wrote:
>> Q. Are both cards clocked at the same speed?
>> A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
>> be clocked at 33MHz
>
>> If the ARM700+FPA runs at 33MHz, I will be *******MOST******* disappointed.
>
>Pass the valium :(
>We can spend an extra 50 quid, and buy a SLOWER machine - I think i've fallen
>into a parallel universe.
>
Well that answer from Chris Cox also is dissapointing to me. In short I now
think that I will not want to upgrade my 610 for some time. The 700 + FPA
would be faster than the 710 only if the software does lots of Floating point.
At present very little software does so the 710 will be faster. If I buy
another machine at the time they are ready I suppose I would opt for the
700+FPA but unless or until then I'm going to sit it out for while and
see which software products make use of FP.
Mike Clark, mr...@cam.ac.uk http://www.path.cam.ac.uk/MikeClark/
--
o/ \\ // |\ ,_ o Mike Clark, C.U. Department of Pathology
<\__,\\ // __o | \ / /\, "a mountain climbing, cycling, skiing,
"> || _`\<,_ |__\ \> | immunology lecturer, antibody engineer and
` || (_)/ (_) | \corn computer user" (Clan No. 1007)
Ale Terlevich
>Q. Are both cards clocked at the same speed?
>
>A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> be clocked at 33MHz
Oh dear.
Thorsten Seitz
> limit specified. The upgrade from the standard 610 to a basic 40MHz 710
> card was quoted at #100
>
> [...]
>
>A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> be clocked at 33MHz
>
>Chris Cox
>Enthusiasts Sales and Marketing Manager
Yesterday I had a look at the information folder Acorn Deutschland had
sent me quite some time ago. It contains the datasheets (translated
into german), which are on ftp://acorn.co.uk, IIRC.
On datasheet 1 (1. edition September 1994 APP 1025) is this nice
little table with the guaranteed processor update prices.
THERE IS NO MENTION OF AN *ARM710* --- instead of a >=40MHz ARM700
and a >40MHz ARM700.
So, why will the ARM700 suddenly be clocked AT ONLY 33MHz ??
What we need is an ARM700 @ 66MHz with FPA11 @ 33MHz !
(And even a FPA11 @ 33MHz is only about as fast as a 486DX2/66.
The new Performa 5200 from Apple (PPC603 @ 75 MHz) is about SIX times
as fast. (Double precision, as indicated by the MFLOPS(1)...MFLOPS(4)
benchmarks.) Single precision (sufficient for most graphics, I think)
will widen the gap.)
So I ask again, what are the current states of the ARM800 and the
StrongARM ?
-Thorsten
Thorsten Seitz
> [...]
>Now, I gave up waiting for an RPC with ARM700 + FPA11 because of all the
>delays, the cost and the prospect of only incremental performance
>improvements.
>
>It seems I was right, as a 33MHz ARM700 + FPA11 seems unlikely to develop
>significantly more Mega-flops than the old 25MHz combination.
Same case for me.
Seems as if I will get a Pentium with NeXTStep instead of a RiscPC
or one of the new PowerMacs.
-Thorsten
Richard West
|> >
|> > be clocked at 33MHz
|>
Whoops.
|> According to the Acorn Processor Technology roadmap a 40MHz ARM 710
|> should be around 40% faster than a 30MHz ARM 610. A 33MHz ARM 700 should
|> be about 15% faster than an ARM 610.
Forty percent? Does it come with a speed harness, some sort of safety net for
those not used to the power? Hmm.
Personally, I'm going to spend my cash on something else.
|>
|> The ARM 800 looks way off: Chris implies that a prototype chip has not
|> yet been made. We shall probably see a faster (66MHz) ARM 700 first
|> (possibly in the RiscPC 700?).
|>
|> Could a technophile please enlighten me as to why achieving high clock
|> speeds for ARM chips is apparently so difficult? Is it because:
|>
|> a) ARM don't have access to 0.5 micron fabrication?
|> b) transferring to a smaller die size is technically very difficult?
|> c) transferring to a smaller die size is very expensive?
|> d) chips running 4x faster than the bus would require faster memory
|> e.g. secondary cache on the processor card?
|> e) something else?
|>
|>
Yes, I'd like an answer to that too...
Steve Pringle
> The upgrade prices are guaranteed (subject to RPI) there was no time
> limit specified. The upgrade from the standard 610 to a basic 40MHz 710
> card was quoted at #100
8<---SNIP
> The FPA prices have not yet been set, but should be around #50 more than
> the 710 card price
>
As I understand it from the original specs issued at the last Harrogate Acorn User Show where tthe RiscPC was launched, the upgrades available (with guaranteed upgrade prices) were as follows:-
30MHz 610 -> >30MHz 610 75ukp
30MHz 610 -> >=40MHz 700 100ukp
30Mhz 610 -> >40MHz 700 125ukp
etc.
No mention was made of the 710 processor (a 700 data sheet was included with the specs, as was an FPA11 data sheet). The offer clearly didn't relate to the 700 series, as the processor part numbers were quoted explicitly (i.e. 610 rather than 600 series, 700 rather than 700 series). Presumably the 710 has replaced the fast 610 option. I would therefore expect to be able to buy a 40Mhz 700 card for 100ukp + RPI (or am I missing something?).
> A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> be clocked at 33MHz
>
This is ***very*** disappointing :(. What we all really want is a 66Mhz 700 card with 33Mhz FPA (as a minimum) for 125ukp +RPI - as promised! The small increase in performance of the 700 cards on offer probably means it isn't worth upgrading until something faster comes along (although people who desperately need the FPU might consider a 33MHz 700 card for 100ukp).
What's the problem, can't ARM Ltd deliver the necessary fast clocked processors? If the chip is available in the necessary clock ratings I can't see a problem in producing a card that utilises them. Can we have some answers please.
--
Steve
-----
Steve Pringle
spri...@latrigg.demon.co.uk
Thilo Manske
>The ARM700 is noticeably more expensive to produce than the ARM710
>simply because it has more pins, which pushes up the packaging cost.
The package ist just plastic, copper and tin (AFAIK) - it doesn't even
cost a pound...
--
------------------------------------------------------------------------------
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.
Thilo Manske
>In message <1995Apr19.1...@leeds.ac.uk> D I Simms wrote:
>> I seriously think Acorn have missed the general mood of the Acorn community.
>> Who really wants a 40Mhz ARM710.
It would be really stupid to buy a 40MHz ARM710 for more than 40 Pounds or 100 DM.
This small speed increase (25%?) isn't worh the money (IMHO), especially If you
can clock your ARM610 up.
>>
>> If 40Mhz is the highest spec chip they can produce I think I will wait until
>> the ARM800/810 or the StrongARM is available.
Or go and buy a P6 or PowerPC-Computer if you need high CPU speeds, because Acorns
are sooo slow :-( - even a 66MHz ARM700 with an 33MHz ARM11 will be slower than
a 75MHz Pentium (in terms of raw erfomance, don't look at windows ;-) )
and there are 120MHz ones available today, the Computers with P6 in a few month
- even before the ARM700 (wich should be available since last october), I bet.
>Likewise --- for what appears to be a (relatively) small speed increase, even
>with the slightly faster clocked ARM710, it hardly seems worth paying for the
>upgrade. I will almost certainly wait for an ARM8xx (with some FP support) or
>StrongARM based card. Buying something like the cards mentioned by Acorn
>would be just a stop-gap.
Yes, but if you wait for a StrongARM you might die of old age... if Acorn will
ever release such a card - In the moment their "technology roadmap" stops with
an ARM800 (they said "available in the 1995s") IIRC.
Ian Griffiths
: Peter Burwood (p...@arcangel.demon.co.uk) wrote:
: : So, here is the challenge. [challenge to compile C for benchmark deleted]
: Your challenge would yield irrelevant figures, anyway, since compilation
: has a significant disk I/O component (it is not as processor intensive
: as you seem to think), which is invariant with CPU.
Disk buffering may not be all that hot on RiscOS, but it's not completely
absent is it? I know there have been problems (e.g. losing the contents
of your disk) when it's enabled in the past, but I assume that this has
been fixed to some extent.
The IO component should not be all that significant: the volume of data
to be read in is not so huge compared to that found in, say, graphics-based
applications, particularly video. You're probably not going to want to
read more than a megabyte of source in or so, and people have told me before
on this group that you can easily get 4Mb a second down that Acorn IO
channel, so I guess it shouldn't require more than a few seconds to read
in the whole lot. If you're that worried, compile it out of a RAM disk.
(Data being written shouldn't matter. Write back is supposed to be
supported to, or write behind as I think Acorn call it. As I say though,
use a RAM disk.)
On my machine the compiler seems to sit around for a while just thinking
fairly often when doing builds. Optimising compilation is recognised
as being fairly processor intensive for non-trivial programs.
Matthew Hambley
In article <19950419....@grafton.demon.co.uk>, jmo...@grafton.demon.co.uk
(John Morley) writes:
|>In message <3n06s0$8...@doc.armltd.co.uk> Clive Jones wrote:
|>
|>> The ARM700 is noticeably more expensive to produce than the ARM710
|>> simply because it has more pins, which pushes up the packaging cost.
|>
|>Isn't it about time you integrated the FPA onto the same die as the ARM,
|>then. Compared with other processor designs ARM floating point performance
|>is embarrassing.
I think No. I like the idea of being able to buy different
configurations of processor for different user needs. Home useres can have
ARM700 and FPA11 where as "power users" can have ARM800 and FPA20 (theoreticaly
of corse). Education might find it useful to have ARM800 and FPA11. I think
it is about time that users had to specify that they *don't* want FPA though.
------------------------------------------------------------------------------
"And the moral of the story is: Stick with what you've got because
basicly, I'm fantastic" - Holly
------------------------------------------------------------------------------
Matthew Hambly, nutter by Royal Appointment | "It doesn't matter if the cat is
term time : cee...@caledonia.hw.ac.uk | black or white as long as it's
other times : MHam...@arcade.demon.co.uk | been cooked in a Wok."
http://www.cee.hw.ac.uk/~ceemah | - K. Thomas
------------------------------------------------------------------------------
Clive...@armltd.co.uk
>Clive...@armltd.co.uk writes:
>>The ARM700 is noticeably more expensive to produce than the ARM710
>>simply because it has more pins, which pushes up the packaging cost.
>
>cost a pound...
Don't forget the gold wires, and the solder, and the machine time in
bonding out the pads to the pins.
--Clive.
(Disclaimer: I wouldn't believe a word of this if I were you...)
Thorsten Seitz
>
>I may get shot down in flames here :-)
Quite true :-)
>As I recall the situation, the clock speed of an ARM processor does not give
>a good comparator for performance when matched against other processors.
True. Take floating point for example: At the SAME clock speed a
PPC601 is about THREE times as fast as a FPA11. And this is not peak
performance (which would reach a factor of about ten for matrix
multiplication, for example). (This is based on the MFLOPS(1)...MFLOPS(4)
benchmarks and on a comparison of both architectures.)
>The assumption that a 40MHz 710 will only perform marginally quicker than an
>ARM3 (33MHz) based A5000 is not sustainable in my view.
But maybe the assumption that a 33MHz ARM700 will only perform
marginally quicker than an ARM610 based RiscPC will be sustainable
even in your (rose coloured, as ever) view??
>I am impressed that POVray performance is now deemed to be an accurate
>benchmark for comparisons to be made. Shame I don't use it :-)
POVray is surely not an ACCURATE benchmark, but concerning floating
point seems to be the only choice for comparisons, because other
programs mostly do not use the FPEmulator.
Shame you don't use floating point arithmetics.
Might be useful in some programs (spread sheets, vector graphics, 3d
graphics, scientific simulations ...).
>Steve Holroyd
who never dares to criticize Acorn. (Maybe because of
those rose-coloured glasses).
-Thorsten
Ian Griffiths
: POVray is surely not an ACCURATE benchmark, but concerning floating
: point seems to be the only choice for comparisons, because other
: programs mostly do not use the FPEmulator.
If you intend to use your computer to generate ray traced pictures using
PovRay, then surely it is the ONLY accurate benchmark.
Thilo Manske
>In article <D7AL7...@festival.ed.ac.uk>,
> a...@roe.ac.uk (Ale Terlevich) wrote:
>> >Q. Are both cards clocked at the same speed?
>> >
>> >A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
>> > be clocked at 33MHz
>>
>> Oh dear.
Yepp :-(
>I may get shot down in flames here :-)
Let's see, what comes ;-)
>As I recall the situation, the clock speed of an ARM processor does not give
>a good comparator for performance when matched against other processors. I
>am sure that some of the responses to the news about the clock speeds of the
>ARM 700/710 have been based on this fallacy.
But you can compare a 40MHz ARM710 to a 30MHZ (or 33MHz ;-) ARM610 - and it's
not much faster; the ARM7 core ist nearlly the same as the ARM6 so that the
MIPS or Dhrystones per MHz figures of the ARM6 and ARM7 are almost identical
IIRC.
And the bigger cache (8KB vs 4KB IDC ), larger TLB (IIRC twice as many entries)
and improved Write Buffers (4 vs 2 Adress buffers) doesn't give you much inprove-
ment at this external to internal ratio as used in the RiscPC: The memory system
is so fast, that an (so slow) 30MHz ARM610 doesn't use much of the bandwidth
(IMHO - of course.).
>The assumption that a 40MHz 710 will only perform marginally quicker than an
>ARM3 (33MHz) based A5000 is not sustainable in my view.
Yes, since you can't plug an ARM 710-Card in the A5000.
>I am impressed that POVray performance is now deemed to be an accurate
>benchmark for comparisons to be made. Shame I don't use it :-)
Don't trust in ANY benchmark results! Especially if they are released
by Intel or Apple ;-)
M J Ebourne
Steve Holroyd <s...@holsoft.demon.co.uk> wrote:
>As I recall the situation, the clock speed of an ARM processor does not give
>a good comparator for performance when matched against other processors. I
>am sure that some of the responses to the news about the clock speeds of the
>ARM 700/710 have been based on this fallacy.
>
>ARM3 (33MHz) based A5000 is not sustainable in my view.
You say that clock speed is not a good indicator of the ARM's power
when compared to other processors. Very true. But this is the SAME
processor (give or take a few k of cache). Hence clock speed is
one of the more accurate measures. The 33MHz ARM700 is not going
to be usefully faster for normal desktop work than the ARM610.
So it's 150 quid for an FPA - great if you don't do anything but
FP.
As to the 40MHz 710 then that's very poor too. Why not 50MHz like the
test versions??? And what about the high performance one promised
only a year ago? Looks like I've joined the growing group of people
with 100 quid to spare then. :-(
--
Martin Ebourne `Comparing things with PCs makes
Electronics, everything look better!' - Sergio Monesi
Southampton University Email: m...@soton.ac.uk
Ale Terlevich
>In article <D7AL7...@festival.ed.ac.uk>,
> a...@roe.ac.uk (Ale Terlevich) wrote:
>
>>
>> >Q. Are both cards clocked at the same speed?
>> >
>> >A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
>> > be clocked at 33MHz
>>
>>
>> Oh dear.
>>
>>
>
>
>a good comparator for performance when matched against other processors. I
True.
>am sure that some of the responses to the news about the clock speeds of the
>ARM 700/710 have been based on this fallacy.
Nope.
>
>The assumption that a 40MHz 710 will only perform marginally quicker than an
>ARM3 (33MHz) based A5000 is not sustainable in my view.
Unfortunately though it's true. However, the A5000 will be crippled
in high res modes by it's lack of VRAM and here the RiscPC would perform
significantly faster.
I would also imagine that a RiscPC with an ARM700 would be faster
than with an equivalent ARM600 if it was also running a PC card due to
the reduced load on the bus (it has a bigger cache).
>
>I am impressed that POVray performance is now deemed to be an accurate
>benchmark for comparisons to be made. Shame I don't use it :-)
>
I just thought I'd test this out, so I compiled povray on a Linux box
with an i486DX2/50 (note, just 50) using gcc2.6.3 and -O (optimised).
It runs 25 times quicker than my RiscPC (Arm310 @ 30MHz) on which I compiled
povray using Acorn C v 4 ( somewhat quicker than gcc2.4.2 or whatever the latest
RiscOS GCC is)
I don't think that the FPU is going to increase the speed of Povray by a factor
of 25, but if anyone thinks differently, please feel free to prove me wrong.
btw. timings were
POV-Ray Options in effect: +v1 +fd +mb25 -q9 -w80 -h60 -s1 -e60
-k0.000 -mv2.0 -ipovscn/scenes/level3/wg5.pov -odata.dis -lpovdoc/include
Linux PC : 73 seconds
RiscPC : 18000 seconds (30 mins) ( approx )
Malcolm Lithgow
: In article <19950419....@sherwood.demon.co.uk> tr...@sherwood.demon.co.uk (Ian Trimble) writes:
: > According to the Acorn Processor Technology roadmap a 40MHz ARM 710
: > should be around 40% faster than a 30MHz ARM 610. A 33MHz ARM 700 should
: > be about 15% faster than an ARM 610.
: I agree, those are the figures suggested by the graph in the technical
: specification flyers that were distributed by Acorn. However, those are
: for the dhrystone benchmark which is greatly effected by cache size.
: What I, and dare I say everyone else here, want to know is what the
: typical increase will be for processor intensive applications, such as a
: compilation of a large system.
: So, here is the challenge. [challenge to compile C for benchmark deleted]
Since the dhrystone test already fits in the cache of the 610, the 40%
figure (if it is accurate) is independant of cache size increase. In
that case, you may expect real applications (like DTP, photo-retouching,
spreadsheets, etc.) to have an even greater speed increase, since they
WILL benefit from the increased cache size.
Your challenge would yield irrelevant figures, anyway, since compilation
has a significant disk I/O component (it is not as processor intensive
as you seem to think), which is invariant with CPU.
-Malcolm lit...@uslp.usl.com
Torben AEgidius Mogensen
>In article <3n2muh$2...@acorn.acorn.co.uk>, cc...@acorn.co.uk (Chris Cox) wrote:
>> The upgrade prices are guaranteed (subject to RPI) there was no time
>> limit specified. The upgrade from the standard 610 to a basic 40MHz 710
>> card was quoted at #100
>> The FPA prices have not yet been set, but should be around #50 more than
>> the 710 card price
>>
>As I understand it from the original specs issued at the last
>Harrogate Acorn User Show where tthe RiscPC was launched, the upgrades
>available (with guaranteed upgrade prices) were as follows:-
>30MHz 610 -> >30MHz 610 75ukp
>30MHz 610 -> >=40MHz 700 100ukp
>30Mhz 610 -> >40MHz 700 125ukp
>etc.
>> A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
>> be clocked at 33MHz
>This is ***very*** disappointing :(. What we all really want is a
>66Mhz 700 card with 33Mhz FPA (as a minimum) for 125ukp +RPI - as
>promised!
That was not what was promised. The upgrade chart does not mention
inclusion of an FPA, nor does it mention 66MHz (just >40MHz, which may
be 41MHz). The problem is that neither of the to-be-released cards fit
the specifications from the chart. The ARM700+FPA card is slower than
the ARM700 options in the chart, and includes an FPA. The ARM710 card
is (as stated) not an ARM700 card, so it doesn't fit the ">=40MHz
ARM700" line exactly. Since no dates were given for the availability
of the upgrades in the chart, it would be perfectly valid for Acorn to
postpone the release of these upgrades indefinitely, in the meantime
selling upgrades with different specifications for different prices.
While I too am dissapointed with the specifications of the new cards
(in particular the ARM700+FPA card), I think Acorn have done the best
they reasonably could. I'm sure they are as frustrated as most of the
people here.
Torben Mogensen (tor...@diku.dk)
M J Ebourne
Olly Betts <ol...@mantis.co.uk> wrote:
>The folks round here are mostly techies, and the sort of people to buy
>new-fangled whizz-bang gadgets (or maybe it's just me). So we're probably a
No, I think you speak for most of us. :-)
>fairer sample of the sort of people who would rush out and buy a faster
>processor card. We're the early demand, which is what Chris Cox was trying
>to judge (as I read it anyway).
Yep, I've had my 100 quid for a decent ARM700 card sitting in the bank
since Xmas. The idea was that I'd buy one ASAP. Looks like I've gotta
think of something else to do with the money though - 100 quid for a
small integer only improvement would be gross mismanagement of my
funds. :-(
If we're not going to be buying them en masse, I wonder who is?
>If I were to upgrade, I'd probably buy an ARM710 if the ARM700 was clocked
>appreciably lower, and hence slower in general use. Whether I upgrade at
>all depends on how much faster it is than an ARM610 and what colour my bank
>balance is at the time.
The minimum I think I'd go for would be 100 quid for a ~60MHz ARM710
or 100 quid for a 50MHz ARM700 with a >=25MHz FPA socket. I don't
need another FPA, I've already got one... IIRC that is largely what
they promised us a year ago.
One worrying point - they said the 700 would be towards the end
of last year - ie about 6-7 months time from when they announced it.
Here we are a year later with it still looking a month or so away.
That's a factor of 2. They said the ARM800 would be towards the end of
this year. ie. about 18 months from the announcement. If you also
apply the *2 factor to that (this *2 factor is very accurate it
seems - just look at the PC card!), then the ARM800 is still
2 years away. Given that Chris Cox said (IIRC) that they are
still playing around with it at chip level, this not only
sounds plausible, but worries me immensely...
Oh, and while I think about it, there's another question I haven't
seen answered yet - what happens to the old processor cards?
Do we get to keep them? I'd certainly hope so at the quoted
price/performance ratios. In which case, what's happened to the
Simtec multiprocessor expansion card which was supposed to be
in existence 6 months ago? Was it stillborn? If not and I
could keep the ARM610 then I might just be persuaded for one
of those ARM700/710 things. (Yes I know RISC OS wouldn't benefit
- I don't care, I'm a hacker, I just wanna play with it. :-))
Tom Hughes
> That was not what was promised. The upgrade chart does not mention
> inclusion of an FPA, nor does it mention 66MHz (just >40MHz, which may
> be 41MHz). The problem is that neither of the to-be-released cards fit
> the specifications from the chart. The ARM700+FPA card is slower than
> the ARM700 options in the chart, and includes an FPA. The ARM710 card
> is (as stated) not an ARM700 card, so it doesn't fit the ">=40MHz
> ARM700" line exactly. Since no dates were given for the availability
> of the upgrades in the chart, it would be perfectly valid for Acorn to
> postpone the release of these upgrades indefinitely, in the meantime
> selling upgrades with different specifications for different prices.
What the data sheet said was a >= 40MHz ARM700 for GBP100, or >40MHz
for GBP125. This is obviously daft, as the first one totally covers
the second, so I have been assuming that is was a misprunt and should
have read <=40MHz and >40MHz, which would at least make sense.
Tom
--
Tom Hughes (t...@compton.demon.co.uk)
5 Lampits, Hoddesdon, Herts EN11 8EH
... CONgress (n) - Opposite of PROgress
Peter Burwood
> Peter Burwood (p...@arcangel.demon.co.uk) wrote:
>
> : [dhrystone performance on 30Mhz 610 vs 40 MHz 710. What will the real
> : performance be like ?]
>
> : So, here is the challenge. [challenge to compile C for benchmark deleted]
>
> Since the dhrystone test already fits in the cache of the 610, the 40%
> figure (if it is accurate) is independant of cache size increase. In
> that case, you may expect real applications (like DTP, photo-retouching,
> spreadsheets, etc.) to have an even greater speed increase, since they
> WILL benefit from the increased cache size.
Partly. I didn't realise that dhrystone fitted within the 4KB cache on
the 610, but was pretty sure it would fit in the 8KB on the 710. I'm
well aware that the cache will help larger applications and what I want
to know is how much the larger cache and TLB, changes to the write
buffer and the faster clocked CPU will affect applications when the main
memory subsystem is still running at 16MHz.
> Your challenge would yield irrelevant figures, anyway, since compilation
> has a significant disk I/O component (it is not as processor intensive
> as you seem to think), which is invariant with CPU.
I disagree. I've some files which take about 10 minutes each to compile
on a RiscPC and the disc I/O component is extremely small, i.e.,
certainly less than 1%. Also, something like a compilation is possibly
more representative of total system performance than DTP and is a lot
easier to benchmark. Ideally, Acorn would produce SPECmarks for the
systems, but that entails a lot of money that Acorn are better spending
elsewhere.
As I said above, the biggest concern is that the memory system is still
running at 16Mhz, so every time the 710 access main memory it's going to
be running at the same speed as the 610. Though main memory accesses
will obviously occur less frequently - but by how much. The dhrystone
test does not give a proper indication of system performance, which is
what we want and that is the point to my original challenge.
Mike James
cc...@acorn.co.uk (Chris Cox) wrote:
> Please could you help us to get it right!
>
> At the launch of the RiscPC we made a commitment to rolling program of
> processor upgrades. To prepare for the release of the 700 upgrade cards
> during the second quarter (ish) I want to make sure that we get the right
> number and mixture. So I am looking for some indication of how many existing
> users feel that they will upgrade to a new card, and whether it will be to
> the 710 integer only card, or the higher cost 700 card with the FPA
> (Floating Point Accelerator)
>
> Details about the different processors were included in the original RiscPC
> datasheets, and can be obtained from the ftp site in
> pub/documents/products/RiscPC/technotes where they are available either
> plain or sparkived.
>
> I look forward to reading your comments.
>
> Chris Cox
> Enthusiasts Sales and Marketing Manager
>
I will only be upgrading the CPU after kitting out with more goodies like
CD-ROM and RAM. As the RISC PC has cost so much it will be some time before
I upgrade.
If I was upgrading I think I would go for
Integer >> 2x faster performance with possibly the option of an FPA in a
socket beside the CPU for say #20 extra (for the empty socket and the extra
ARM pins)
Soldered joints that dont have solder paste between the pins on the CPU card.
Thats what the _replacement_ ARM610 card came with.
I dont think the ARM700/710 is fast enough to tempt.
The difference in speed between ARM2 and 36MHz ARM3 was enough to make me pay
#150 for 10 months happy use (before it fried itself and I got the RISC PC)
Really I think I will wait for 4 times integer speed improvement before
upgrading. This means a StrongARM or a very quick ARM810 ....
Mike.
--
---------------------------------
-- Mike James G6IXE --
-- Using an Acorn RISC PC 600 -- My other Computer's a ... 6800 .. 6809 ..
-- a British, ARM Powered -- ZX81 .. Jupiter Ace .. BBC B .. A310 ..
-- Non-Intel, 32 bit machine -- A3010 .. Psion Series 3 ..
---------------------------------
139 Error undeclared name, inventing 'extern int Curr_Eng'
Tom Hughes
> The minimum I think I'd go for would be 100 quid for a ~60MHz ARM710
> or 100 quid for a 50MHz ARM700 with a >=25MHz FPA socket. I don't
> need another FPA, I've already got one... IIRC that is largely what
> they promised us a year ago.
As the price on the original data sheet for a >40MHz ARM700 was
GBP125, you would not be likely to get a 60MHz one for GBP100
Martin :-)
Tom
--
Tom Hughes (t...@compton.demon.co.uk)
5 Lampits, Hoddesdon, Herts EN11 8EH
... Hardware: The part you kick.
Tony Finch
Tom Hughes <t...@compton.demon.co.uk> wrote:
>
> What the data sheet said was a >= 40MHz ARM700 for GBP100, or >40MHz
> for GBP125. This is obviously daft, as the first one totally covers
> the second, so I have been assuming that is was a misprunt and should
> have read <=40MHz and >40MHz, which would at least make sense.
I read it as implying that the minimum upgrade speed would be 40MHz,
and there would be two options, "faster" and "even more fast".
Tony.
--
"You mows it and you rolls it for five hundred years
and then a bunch of bastards walks across it."
Ale Terlevich
>In article <3n33p2$4...@epsilon.qmw.ac.uk>, Henry <en...@dcs.qmw.ac.uk> wrote:
>>> be clocked at 33MHz
>>
>>> If the ARM700+FPA runs at 33MHz, I will be *******MOST******* disappointed.
>>
>>Pass the valium :(
>>We can spend an extra 50 quid, and buy a SLOWER machine - I think i've fallen
>>into a parallel universe.
>>
>
>Well that answer from Chris Cox also is dissapointing to me. In short I now
>think that I will not want to upgrade my 610 for some time. The 700 + FPA
>would be faster than the 710 only if the software does lots of Floating point.
>At present very little software does so the 710 will be faster. If I buy
>another machine at the time they are ready I suppose I would opt for the
>700+FPA but unless or until then I'm going to sit it out for while and
>see which software products make use of FP.
>
Well, I plan to install RiscBSD as soon as it sees the light of day, and run
FP intensive sw on that, so it's still worth the upgrade. Also, the Acorn
processor chart only showed the dhrystone results. As the Arm700 has double the cache
of the Arm610, I'm sure that 'real' applications will see a slightly better
speedup.
Perhaps I should get these rose tinted spectacles permanently welded on, that
way when the StrongARM comes out at 50MHz, I won't be *too* dissapointed!
Ale.
Tony Howat
mr...@cus.cam.ac.uk (Dr M.R. Clark) wrote:
> In article <3n33p2$4...@epsilon.qmw.ac.uk>, Henry <en...@dcs.qmw.ac.uk> wrote:
> >Chris Cox (cc...@acorn.co.uk) wrote:
> >> Q. Are both cards clocked at the same speed?
> >> A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> >> be clocked at 33MHz
> >
> >and Ale Terlevich (a...@roe.ac.uk) wrote:
> >> If the ARM700+FPA runs at 33MHz, I will be *******MOST******* disappointed.
> >
> >Pass the valium :(
> >We can spend an extra 50 quid, and buy a SLOWER machine - I think i've fallen
> >into a parallel universe.
> >
>
> Well that answer from Chris Cox also is dissapointing to me. In short I now
> think that I will not want to upgrade my 610 for some time. The 700 + FPA
> would be faster than the 710 only if the software does lots of Floating point.
> At present very little software does so the 710 will be faster. If I buy
> another machine at the time they are ready I suppose I would opt for the
> 700+FPA but unless or until then I'm going to sit it out for while and
> see which software products make use of FP.
Indeed, I would be quite happy to pay an extra fifty quid for a FPA - but
plain processor performance is my priority. If an FPA means knocking down
the 710's speed I'm not interested in one. Unfortunately that then makes
the cost/advantages ratio of a plain 710 card look rather poor. I'll
stay with my 610 until a card with a substansial speed increase over it
comes along (ie 2x).
Tony.
------------/\/-----------------------------/\/----------------------------
Tony Howat : tho...@xargle.demon.co.uk : Southport, Merseyside U.K.
------------/\/-----------------------------/\/----------------------------
At any time, at any place, our snipers can drop you. Have a nice day.
---------------------------------------------------------------------------
Torben AEgidius Mogensen
>What the data sheet said was a >= 40MHz ARM700 for GBP100, or >40MHz
>for GBP125. This is obviously daft, as the first one totally covers
>the second, so I have been assuming that is was a misprunt and should
>have read <=40MHz and >40MHz, which would at least make sense.
The way I read it was that an ARM700 at >=40MHz would be GBP100 and
an ARM700 faster than the GBP100 board would be GBP125.
Torben Mogensen (tor...@diku.dk)
Paul Boddie
: In message <3n2muh$2...@acorn.acorn.co.uk> Chris Cox wrote:
:
: > Q. Are both cards clocked at the same speed?
: >
: > A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
: > be clocked at 33MHz
: According to the Acorn Processor Technology roadmap a 40MHz ARM 710
: should be around 40% faster than a 30MHz ARM 610. A 33MHz ARM 700 should
: be about 15% faster than an ARM 610.
From a demo I saw last term of the ARM700 + FPA @ 33MHz, floating
point performance was much, much better than the standard ARM610
as you would expect. With ArtWorks files, using the viewer and one of
the car example files, the speed difference was barely noticable -
so don't really expect much of a gain in integer performance.
But the card is definitely worth it for FP performance gains alone.
What about a secondary card so you could have an ARM610 _and_ the
ARM700 + FPA combination.
BTW, I was told that it is not possible to have a 66MHz ARM700 operating
together with a 33MHz FPA.
: The ARM 800 looks way off: Chris implies that a prototype chip has not
: yet been made. We shall probably see a faster (66MHz) ARM 700 first
: (possibly in the RiscPC 700?).
What about the portable?!
--
Paul Boddie ce...@caledonia.hw.ac.uk
Computer Science 3 ce...@clust.hw.ac.uk
Heriot-Watt University ce...@cee.hw.ac.uk
Edinburgh
------------------------------- http://www.cee.hw.ac.uk/~ceepb
Stuart Tyrrell
: In article <1995Apr20....@arbi.Informatik.Uni-Oldenburg.DE> Thilo....@arbi.informatik.uni-oldenburg.de (Thilo Manske) writes:
: >The package ist just plastic, copper and tin (AFAIK) - it doesn't even
: >cost a pound...
: Don't forget the gold wires, and the solder, and the machine time in
: bonding out the pads to the pins.
Thank goodness we don't have to have ceramic packages with porus heatsinks
like _some_ CPU's I could mention - the packaging costs more than a
whole ARM.
Stuart
--------------------------------------+----------------------------------------
Stuart Tyrrell |
s...@sn2.ee.umist.ac.uk | Delete sig?
Room MB/A19F, UMIST, PO Box 88, |
Manchester M60 1QD. Tel 0161-200-4714 | Cancel No Yes Quiet
--------------------------------------+----------------------------------------
Malcolm Lithgow
: Clive...@armltd.co.uk writes:
: >The ARM700 is noticeably more expensive to produce than the ARM710
: >simply because it has more pins, which pushes up the packaging cost.
: The package ist just plastic, copper and tin (AFAIK) - it doesn't even
: cost a pound...
And the chip inside is just melted sand! Why is it so expensive!?
Hey, I just thought of something: a Pentium's package is just plastic,
copper, and tin (actually, there's probably some gold and stuff in there
too), it can't cost more than a few pounds, how come it's hundreds?
Wakey, wakey. In case you haven't noticed, the price of things isn't just
the value of the raw materials.
-Malcolm lit...@uslp.usl.com
Jens H. Ovesen
> > I look forward to reading your comments.
> >
> > Chris Cox
> > Enthusiasts Sales and Marketing Manager
> >
>
> I would definitely buy the 700 FPA card.
So would I. I actually need a FPA to be able to do some of my work at
home.
Jens.
Dave Bass
[munch]
> From a demo I saw last term of the ARM700 + FPA @ 33MHz, floating
> point performance was much, much better than the standard ARM610
> as you would expect. With ArtWorks files, using the viewer and one of
> the car example files, the speed difference was barely noticable -
> so don't really expect much of a gain in integer performance.
>
> But the card is definitely worth it for FP performance gains alone.
> What about a secondary card so you could have an ARM610 _and_ the
> ARM700 + FPA combination.
That's all very well, but how about comparing like with like.
*I* want to know what the relative performance of the A700 + FPA11 @ 33MHz
is compared to the ARM3 + FPA10 @ 25MHz, not how much faster the ARM700 is
than the ARM610.
Chris Cox said 3-5 MFLOPS for the new item. Can anybody tell me what the
old one's rating was/is?
--
David Bass, ESTEC/YCV, Postbus 299, 2200 AG Noordwijk, The Netherlands
#include std_disclaimer
Chris Cox
the 700 cards. I hope that the following will help clarify some of the
issues involved.
The ARM700 and ARM710 processors represent a significant improvement over
the current ARM610 processors. They have a higher maximum clock speed and a
number of architectural improvements such as double the size of internal
cache, this means that more of any process can be executed internally
without accessing the (relatively) slow external memory. Other improvements
are an improved write buffer and an enlarged Translation Lookaside Buffer in
the MMU. All of these improvements increase the performance of the system
and deliver more real performance than a simple comparison of clock speeds
would indicate.
For floating point work the FPA11 processor has a peak throughput of up to 5
MFLOPS (Millions of FLoating point Operations Per Second) and
achieves an average throughput in excess of 3 MFLOPS for a range of
calculations. Applications which make use of the RISC OS Floating Point
Emulator will automatically use the FPA chip, increasing floating point
performance by as much as a factor of 50. This will make a huge difference
in the performance of floating point intensive tasks such as ray tracing.
The FPA11 is interfaced to the ARM700 over a simple, high-performance
co-processor bus. The ARM instruction pipeline is mirrored in FPA11 so that
floating point instructions can be executed directly with minimum
communication overhead. Pipelining, concurrent execution units and
speculative execution are all used to improve performance without
significantly impacting power consumption. The ARM700 processor, which has
the external coprocessor bus interface, is currently fabricated on a process
that limits the maximum clock speed to 33MHz.
The ARM710 has been optimised for integer performance, and of the samples
that we have evaluated the fastest ran at 55MHz However there is a big
difference between a prototype based on a sample and a production board.
When we specify the production boards we have to make sure that they will
work under all conditions, and for all devices. This means that we have to
derate the clock speed to allow for a number of different factors:
Fastest ever ARM710 sample seen running in ideal conditions 55MHz
Allow for power drop to 4.75V eg when HDD spins up 0.97
Allow for a max chip temp of 70 deg C 0.9
(30deg ambient +15deg inside case +25deg inside chip)
Process spread from the fastest part 0.825
Multiplying these together gives the specification of 40MHz
These are the figures that we use, based on our experience of a number of
different fabricators. As you can see from the above figures it is
perfectly possible that with the right crystal, a particular board in a
particular system might run at 50MHz to 55MHz, but this cannot be
guaranteed.
Chris Cox
Enthusiasts Sales and Marketing Manager
(Please send Clan related mail to cl...@acorn.co.uk thanks)
Ray Briddock
DP> speed increase, even with the slightly faster clocked
DP> ARM710, it hardly seems worth paying for the upgrade. I will
DP> almost certainly wait for an ARM8xx (with some FP support)
DP> or StrongARM based card. Buying something like the cards
DP> mentioned by Acorn would be just a stop-gap.
A 40% speed increase is better than the spped increase of the first gen.
Pentiums over the 486's that were available.
I think Acorn have done a sterling job in getting the ARM7 based processors
out one year after a major redesign like the RPC was.
Intels average appears to be about 24 months between generations. If
Acorn/ARM can maintain the 12 month average then it won't be long before we
can look forward to true speed superiority again.
PLUS ... I don't need to upgrade my entire machine to take advantage of it.
Ray - <Gis a job Acorn :-) .... >
--
___ ___ ___ ___ ___ ___
| / \/ \/ \/ \/ \/ \ | The Archimedes Bulletin Board in London |
| A R C A D E | +44(0)181 654 2212 (4 lines) Croydon, UK|
| Email & News @arcade.demon.co.uk | +44(0)181 655 4412 Fidonet#2:254/27.0 |
Ray Briddock
IR> I'm hanging around in the hope that multiple processors will
IR> arrive one day.
I am sure they will. I am sure I will be able to install them in my RPC.
Until that day I have the advantage of using possibly the best OS in the
west running on the machine most suited to show it at its best.
Ray
Ray Briddock
PB> then I want it available as an upgrade now (assuming of
PB> course the RiscPC 700 is going to appear shortly). If Acorn
PB> stuff us RiscPC 610 owners by not allowing us to upgrade to
PB> similar spec machines then there are going to a lot of
PB> pissed people.
But Acorn have always done this. A305/A310 not a patch on A4x0/1 series and
difficult upgrade path. I understood this when choosing to buy a first
generation RPC because I understood that if I didn't then Acorn just might
noy be around to offer any newer better machines.
Yer pays yer money .....
Ray Briddock
CC> A. The 710 card will be clocked at 40MHz and both the 700
CC> and FPA will be clocked at 33MHz
In that case I'll be having a 710. There is nothing I currently do which
requires hardware FP.
Thorsten Seitz
> In article <1995Apr20....@arbi.Informatik.Uni-Oldenburg.DE> Thilo....@arbi.informatik.uni-oldenburg.de (Thilo Manske) writes:
> >>The ARM700 is noticeably more expensive to produce than the ARM710
> >>simply because it has more pins, which pushes up the packaging cost.
> >
>
> Don't forget the gold wires, [stuff deleted]
>
> --Clive.
I ask again:
What are the prices of an ARM710 and an ARM700 ?
$20 or $30 ?? IIRC not much more than that.
So, what's the point ?
(the card costs L125 (may I ask, why?))
-Thorsten
Thorsten Seitz
*I* want to know what the relative performance of the A700 + FPA11 @ 33MHz
> is compared to the ARM3 + FPA10 @ 25MHz, not how much faster the ARM700 is
> than the ARM610.
>
> Chris Cox said 3-5 MFLOPS for the new item. Can anybody tell me what the
> old one's rating was/is?
The FPA10 and FPA11 are *identical*. The only difference is, that the FPA11
can be clocked at higher rates.
-Thorsten
Jeremy Bradley
functionality. It strikes me that at some stage Acorn have to bite
the bullet and say:
"From now on RPCn00+ will have FPA functionality as standard_"
Until then FPAs will remain of minority interest to
so-called "power-users". The reason being is that while an FPA is an
optional extra at best, software developers are not going to make use
of it - however if it were to be made a standard fitting then you
would probably see a big increase in performance of subsequently
released packages - because developers would be assured that every
machine after a certain point would have this feature.
Just a thought.
--Jeremy.
Ian Rawlings
cc...@acorn.co.uk (Chris Cox) wrote:
>Q. Are both cards clocked at the same speed?
>
>A. The 710 card will be clocked at 40MHz and both the 700 and FPA will
> be clocked at 33MHz
33MHz? Is there any speed advantage over the 610 at this speed? This
sounds pretty slow, so it looks like I'll be waiting for a lot longer..
Any news on faster clock rate cards, eg 60MHz?
--
Opinions? I am a Fortean!
Ian Rawlings, i...@daffodif.demon.co.uk - England.
Simon Burrows
> Please could you help us to get it right!
>
> At the launch of the RiscPC
Shouldn't that be written Risc PC?
^
We need to know what to embroider on the sleeves of our anoraks.. ;->
--
Simon Burrows - s...@nucleus.demon.co.uk
S.Bu...@lc03.aud.london.coopers.colybrand.gold-400.gb
Thorsten Seitz
>The ARM700 and ARM710 processors represent a significant improvement over
>the current ARM610 processors. They have a higher maximum clock speed and a
Obviously not the ARM700.
>For floating point work the FPA11 processor has a peak throughput of up to 5
>MFLOPS (Millions of FLoating point Operations Per Second) and
>achieves an average throughput in excess of 3 MFLOPS for a range of
>calculations. Applications which make use of the RISC OS Floating Point
>Emulator will automatically use the FPA chip, increasing floating point
>performance by as much as a factor of 50. This will make a huge difference
>in the performance of floating point intensive tasks such as ray tracing.
Agreed, but compared to other systems (Pentium, PowerPC) the FPA11 is
slower by a factor of about 6, for some purposes by a factor of 10 or
more.
Therefore I would like to know what we can expect in the future.
Is there a succesor to the FPA11 in development ?
Will a FPA be integrated on the ARM800, on the StrongARM ?
>significantly impacting power consumption. The ARM700 processor, which has
>the external coprocessor bus interface, is currently fabricated on a process
>that limits the maximum clock speed to 33MHz.
Is there anything being done to move it to a fabrication process with
a higher limit on the clock speed ?
If so, what limit would this be, and when could we expect cards with
it ?
-Thorsten
Ian Trimble
>
> The ARM700 and ARM710 processors represent a significant improvement over
> the current ARM610 processors. They have a higher maximum clock speed and a
> number of architectural improvements such as double the size of internal
> cache, this means that more of any process can be executed internally
> without accessing the (relatively) slow external memory. Other improvements
> are an improved write buffer and an enlarged Translation Lookaside Buffer in
> the MMU. All of these improvements increase the performance of the system
> and deliver more real performance than a simple comparison of clock speeds
> would indicate.
We should congratulate Chris for being so candid. It's hard to say how
much faster a machine will "feel" with the new cards.
RiscPC owners upgrading will probably find the upgrade similar to the
change from a 25MHz to a 33MHz A5000 (ARM710) or adding an FPA10 to a
25MHz A5000 (ARM700 + FPA11).
A5000 owners (with video cards) upgrading from a 25MHZ ARM3 + FPA to an
ARM700 machine or from a 33MHz ARM3 to an ARM710 machine will probably
notice a doubling of performance.
There seems to be a feeling on c.s.a. that Acorn has stalled.
I'm inclined to think they have diverted resources from further
development of the RiscPC to their next-generation machine, having
realised that the RiscPC is no Power-PC beater. Peter Bondar hinted as
much at last year's Harrogate presentation.
Chris states that the ARM700 is currently fabricated on a process that
limits the maximum clock speed to 33MHz. We know that Acorn / ARM
have established a link with Digital to move to 0.35m fabrication, but
I imagine that moving to a new fabrication process is not trivial
(the latest Pentium 120 is the first Pentium to use 0.35m fabrication;
2 years after the first release).
The next significant ARM chip is likely to use 0.35m fabrication and to
have an ARM800 core with its fancy new (asynchronous?) tricks and
integral floating point unit running at some outrageous (200+MHz?)
clock speed. It will probably have an enormous on board cache to cope
with the flow of data. (Alphas use large amounts of interlaced secondary
cache; the new Intel P6 has 256k cache on board!)
I can't see how the current RiscPC architecture could keep up with this
sort of leap in CPU performance to allow a "plug-in card" upgrade.
Implementation of the new processor will probably require considerable
parallel development of the bus and operating system.
Apple may inject a sense of urgency into the project: they are keen to
enhance their flagging Newton and I'm sure they don't want to re-write
their software from scratch.
Given the amount of work involved I'd be surprised if we don't have to
wait a year or more for Acorn's next release, but it should be something
special.
Ever the optimist.
__ ____
/ _ _ / _ ' _ _ / / _
___/__(_|_| )___/_/ _(_| ) )_/_)_(_(/___________________________________
Dr Ian Trimble email: tr...@sherwood.demon.co.uk
Sherwood Health Centre
Elmswood Gardens Tel: +44 115 962 4516
Nottingham NG5 4AD Fax: +44 115 985 7899
________________________________________________________________________
Thorsten Seitz
Very right!
The hopefully-just-around-the-corner new Risc PC 700 should introduce
this standard.
-Thorsten
A.W. Garrard
>>
>>I am impressed that POVray performance is now deemed to be an accurate
>>benchmark for comparisons to be made. Shame I don't use it :-)
>>
>
>
> I just thought I'd test this out, so I compiled povray on a Linux box
>with an i486DX2/50 (note, just 50) using gcc2.6.3 and -O (optimised).
>
> It runs 25 times quicker than my RiscPC (Arm310 @ 30MHz) on which I compiled
>povray using Acorn C v 4 ( somewhat quicker than
> gcc2.4.2 or whatever the latest
>RiscOS GCC is)
>
> I don't think that the FPU is going to increase the
> speed of Povray by a factor
>of 25, but if anyone thinks differently, please feel free to prove me wrong.
Hmm. Assuming PoVRay uses the FPEmulator (haven't checked), I don't think it
will be far off that. I remember (very) old figures claiming that emulated
floating point instructions could be about 70 times slower than the old
Weitek FPU; the emulator will be faster now, but then so is FPA11. Love to
give you some figures, but since I heard (I think) ARM700 isn't coming until
after ARM710, it may be some time before I can tell. *sulk*
Now, if you want an application that will make a 486DX2 look good and an ARM
look bad, I think something with a tight-ish inner loop that will fit in a
secondary cache, and which uses lots of floating point (note that with FP
emulation, the ARM cache will keep stalling), will do quite well. Since the
heart of a ray-tracer is a small-ish loop (depending on how many options you
use) and a lot of floating point number crunching, you may have just found
why Acorns aren't well known in the rendering market, where Amigas have found
a niche (a 40MHz 68040 is slower on integer than a 66MHz 486-DX2, but blows
it away on floating point).
I'm just waiting for the ARM700 card, and sulking quietly.
Andrew 'oh no he's back' Garrard
A.W. Garrard
>There has been a lot of discussion about the clock speeds and performance of
>the 700 cards. I hope that the following will help clarify some of the
>issues involved.
> The ARM700 processor, which has
>that limits the maximum clock speed to 33MHz.
Ish, give or take the ones at shows running at 40MHz...
>The ARM710 has been optimised for integer performance, and of the samples
>that we have evaluated the fastest ran at 55MHz However there is a big
>difference between a prototype based on a sample and a production board.
>When we specify the production boards we have to make sure that they will
>work under all conditions, and for all devices. This means that we have to
>derate the clock speed to allow for a number of different factors:
>
>Fastest ever ARM710 sample seen running in ideal conditions 55MHz
>
>Allow for power drop to 4.75V eg when HDD spins up 0.97
>
>Allow for a max chip temp of 70 deg C 0.9
>(30deg ambient +15deg inside case +25deg inside chip)
>
>Process spread from the fastest part 0.825
>
>Multiplying these together gives the specification of 40MHz
>
>These are the figures that we use, based on our experience of a number of
>different fabricators. As you can see from the above figures it is
>perfectly possible that with the right crystal, a particular board in a
>particular system might run at 50MHz to 55MHz, but this cannot be
>guaranteed.
I see Acorn are learning from the '33MHz ARM3 in the A540' experience.
Any chance of testing cards at different speeds, and selling the faster
ones at a slightly higher price? That gives 48MHz, and I'd pay another 20ukp
for a fan-boosted heat sink and some power smoothing circuitry. I'm sure
Aleph 1 claimed to have an ARM710 at 50MHz at a recent show. 50Mhz would
give a speed-up of just under x2 on an ARM610, which is about worthwhile.
Now, I would also like some FPA performance; do we know about the multi-
processor card? Assuming Risc OS won't use it for now, does it take 'first'
or 'second' processors? If it copes with ASIC-less processors, it would be
really nice to have an ARM710 at a high clock speed as main processor, with
an ARM700+FPA and PC card (and eventually something nicer, like an alpha)
on the multiprocessor board. Then if I were going to do FP-intensive work,
on Risc OS or something which doesn't utilise the extra processors, I could
swap the ARM cards. This also brings up the point of how much processor
cards cost if you're not replacing an old one (of course, then the multi-
processor board would have to have lots of ASICs, which may prove difficult).
But it would be a nice thought. Is that multiprocessor board still going for
100ukp? I like the 1Mb cache idea, apart from anything else - anyone know
how fast that RAM was? Presumably nippy, or dual-ported, or it wouldn't be
much use as a cache.
Oh well, call me an optimist. One other thing - I was looking at the card
slots, and I can't see how the multiprocessor card is going to fit in unless
the machine has at least two slices. Can anyone enlighten me?
Andrew will-a-285MHz-alpha-overheat-my-ARM610 Garrard