pidd
Posts: 4156
Joined: Fri May 29, 2020 8:29 pm
Location: Wirral, UK

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 3:04 pm

Health warning - you have to wait until next month for part two.

ejolson
Posts: 10219
Joined: Tue Mar 18, 2014 11:47 am

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 3:56 pm

It tells a nice story. I wish these articles on computer history were more careful with their citations as it's not journalistic news with confidential sources.

Did I miss where they were listed?

Heater
Posts: 19234
Joined: Tue Jul 17, 2012 3:02 pm

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 4:12 pm

If you search youtube for "steve furber acorn arm" you can get the story from the horses mouth. There are many interviews with Steve about the early days of Acorn and the development of the ARM. All good stuff.
Slava Ukrayini.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 31817
Joined: Sat Jul 30, 2011 7:41 pm

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 5:35 pm

ejolson wrote:
Sat Sep 24, 2022 3:56 pm
It tells a nice story. I wish these articles on computer history were more careful with their citations as it's not journalistic news with confidential sources.

Did I miss where they were listed?
It all sounds about right to me. Fairly common knowledge.
Principal Software Engineer at Raspberry Pi Ltd.
Working in the Applications Team.

User avatar
MikeDB
Posts: 1367
Joined: Sun Oct 12, 2014 8:27 am

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 8:30 pm

Far better to look at Herman Hauser's postings on what really happened in the early years, and then of Robin Saxby later on. Engineers tend to have a rather narrow view on the whole business picture. ARM didn't succeed because of good engineering, but because of great marketing.

There actually was a better RISC processor doing the rounds at the same time which several semiconductor companies, including Philips and Fujitsu, looked at but the guy wouldn't accept that the engineering was at best 10% of the job and so nobody picked it up in the end.
Always interested in innovative audio startups needing help and investment. Look for InPoSe Ltd or Future Horizons on LinkedIn to find me (same avatar photograph)

Heater
Posts: 19234
Joined: Tue Jul 17, 2012 3:02 pm

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 8:59 pm

MikeDB wrote:
Sat Sep 24, 2022 8:30 pm
ARM didn't succeed because of good engineering, but because of great marketing.
Well, it probably helps if your marketing guys have something good to market. Otherwise the marketing guys are just a fraud selling garbage.
MikeDB wrote:
Sat Sep 24, 2022 8:30 pm
There actually was a better RISC processor doing the rounds at the same time which several semiconductor companies, including Philips and Fujitsu, looked at but the guy wouldn't accept that the engineering was at best 10% of the job and so nobody picked it up in the end.
I'm curious to know what better RISC processor that was. Care to let us know?

I find you notion that "engineering was at best 10% of the job" distasteful and factually incorrect. A modern example is Tesla that has the highest profits per vehicle sold than any other car maker, despite not spending a cent on advertising.

Although perhaps ARM were lucky in that Apple came along to finance ARM for their Newton project. But was that because ARM had the best engineering design available at the time...
Slava Ukrayini.

pidd
Posts: 4156
Joined: Fri May 29, 2020 8:29 pm
Location: Wirral, UK

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 9:05 pm

Heater wrote:
Sat Sep 24, 2022 8:59 pm
A modern example is Tesla .... despite not spending a cent on advertising.
I'm afraid that is factually incorrect.

Heater
Posts: 19234
Joined: Tue Jul 17, 2012 3:02 pm

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 9:32 pm

pidd wrote:
Sat Sep 24, 2022 9:05 pm
Heater wrote:
Sat Sep 24, 2022 8:59 pm
A modern example is Tesla .... despite not spending a cent on advertising.
I'm afraid that is factually incorrect.
Is it? The "facts" coming to me are clearly different than those coming to you. Can you link me to some contrary "facts"?
Slava Ukrayini.

pidd
Posts: 4156
Joined: Fri May 29, 2020 8:29 pm
Location: Wirral, UK

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 9:42 pm

I've seen Tesla adverts

https://www.youtube.com/watch?v=QweNsLesMrM

I would be amazed if they don't have some sort of product placement advertising, Elon was trying to buy an advertising medium called Twitter.

User avatar
rpiMike
Posts: 2230
Joined: Fri Aug 10, 2012 12:38 pm
Location: Cumbria, UK

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 9:56 pm

pidd wrote:
Sat Sep 24, 2022 9:42 pm
I've seen Tesla adverts

https://www.youtube.com/watch?v=QweNsLesMrM

I would be amazed if they don't have some sort of product placement advertising, Elon was trying to buy an advertising medium called Twitter.
I don’t believe what you linked is a Tesla advert.

I’ve never seen an advert and just bought a model Y - which is awesome!

User avatar
lurk101
Posts: 1920
Joined: Mon Jan 27, 2020 2:35 pm
Location: Cumming, GA (US)

Re: Ars Technica article about the birth of the ARM CPU's.

Sat Sep 24, 2022 10:24 pm

Heater wrote:
Sat Sep 24, 2022 8:59 pm
MikeDB wrote:
Sat Sep 24, 2022 8:30 pm
ARM didn't succeed because of good engineering, but because of great marketing.
Well, it probably helps if your marketing guys have something good to market. Otherwise the marketing guys are just a fraud selling garbage.
MikeDB wrote:
Sat Sep 24, 2022 8:30 pm
There actually was a better RISC processor doing the rounds at the same time which several semiconductor companies, including Philips and Fujitsu, looked at but the guy wouldn't accept that the engineering was at best 10% of the job and so nobody picked it up in the end.
I'm curious to know what better RISC processor that was. Care to let us know?
Both MIPS and SPARC are roughly of the same vintage and I'd consider either a better RISC.
I find you notion that "engineering was at best 10% of the job" distasteful and factually incorrect. A modern example is Tesla that has the highest profits per vehicle sold than any other car maker, despite not spending a cent on advertising.
Commanding high profits when you're effectively the single source is not proof of great engineering, in fact the early Teslas were notoriously bug ridden. It was more like pressure cooker engineering. Much of it's success is due to cult of personality. I suspect the 10% figure is merely hyperbole to make a point. In any case how would you quantify accurately anyway?
Although perhaps ARM were lucky in that Apple came along to finance ARM for their Newton project. But was that because ARM had the best engineering design available at the time...
I would argue that technology wasn't the primary factor. I think it mattered more what kind of deal Jobs could swing. Demonstrably the ARM team could do it on the cheap.
The Linux Foundation is like loggers who claim to speak for the trees.

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 31817
Joined: Sat Jul 30, 2011 7:41 pm

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Sep 25, 2022 9:34 am

A good example of a product that sell well with no advertising is .....wait for it....the Raspberry Pi.

We sell on engineering quality and price. Not sure we have ever spent any money on advertising. Unless you count having a stall at a trade show, which we have only done a few times.
Principal Software Engineer at Raspberry Pi Ltd.
Working in the Applications Team.

User avatar
bensimmo
Posts: 5746
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Sep 25, 2022 10:06 am

rpiMike wrote:
Sat Sep 24, 2022 9:56 pm
pidd wrote:
Sat Sep 24, 2022 9:42 pm
I've seen Tesla adverts

https://www.youtube.com/watch?v=QweNsLesMrM

I would be amazed if they don't have some sort of product placement advertising, Elon was trying to buy an advertising medium called Twitter.
I don’t believe what you linked is a Tesla advert.

I’ve never seen an advert and just bought a model Y - which is awesome!
Twitter and You tube are big advert platform for it, not your traditional methods. 'paying influencers and bits and similar' and also on other social platforms.

Sticking one on a 🚀 is also advertising and cost money.

It does not have to be an advert in the traditional sense any more.


You can search YouTube for advert, ad, commercial and tesla.
And of course all the 'reviews'

User avatar
bensimmo
Posts: 5746
Joined: Sun Dec 28, 2014 3:02 pm
Location: East Yorkshire

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Sep 25, 2022 10:09 am

jamesh wrote:
Sun Sep 25, 2022 9:34 am
A good example of a product that sell well with no advertising is .....wait for it....the Raspberry Pi.

We sell on engineering quality and price. Not sure we have ever spent any money on advertising. Unless you count having a stall at a trade show, which we have only done a few times.
'little advertising'
You use social media, people are paid for that job, you give (pre)view samples out, negligible cost, great adverts.

And a publishing wing, one big advert that is paid for ;-) but hopefully/maybe pays for itself?

And an education wing...

The best part, once you've done the initial kick of advertising, in your mags, social media posts, reviews, interviews with videos, etc. It becomes self advertising with retailers doing it for you, education (code clubs, schools, learning resources), books on how to use and of course the community themselves, some even going as far as paying to be a walking advert with tea shirts etc.
(cool ones at that)

A great way to go about it.

User avatar
MikeDB
Posts: 1367
Joined: Sun Oct 12, 2014 8:27 am

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Sep 25, 2022 10:33 am

Heater wrote:
Sat Sep 24, 2022 8:59 pm
MikeDB wrote:
Sat Sep 24, 2022 8:30 pm
ARM didn't succeed because of good engineering, but because of great marketing.
Well, it probably helps if your marketing guys have something good to market. Otherwise the marketing guys are just a fraud selling garbage.
MikeDB wrote:
Sat Sep 24, 2022 8:30 pm
There actually was a better RISC processor doing the rounds at the same time which several semiconductor companies, including Philips and Fujitsu, looked at but the guy wouldn't accept that the engineering was at best 10% of the job and so nobody picked it up in the end.
I'm curious to know what better RISC processor that was. Care to let us know?

I find you notion that "engineering was at best 10% of the job" distasteful and factually incorrect. A modern example is Tesla that has the highest profits per vehicle sold than any other car maker, despite not spending a cent on advertising.

Although perhaps ARM were lucky in that Apple came along to finance ARM for their Newton project. But was that because ARM had the best engineering design available at the time...
If you look back at it, the first ARM processor wasn't that good. But it filled a market niche and Robin Saxby sold it brilliantly into that niche. He gives lots of good anecdotes about it over a beer or three, but I believe the only book with him involved is the one by Daniel Nenni which I think cleanses history a bit.

The better processor is long dead and buried as the marketing, basically the boss's wife phoning people up, was crap. But from our evaluations it did have about a 25% better performance/Watt than the current ARM design at that time. But obviously it wouldn't have a hope against the modern ARM products - ARM engineers have learnt a huge amount about processor design since the early days.

As for being factually incorrect, I'm an engineer as well but once you look at all the effort that goes into marketing and selling a successful product you realise it's usually not the largest time involvement. We always used to moan about this even at HP where we believed our test instruments were always world-leading, which they generally were, and hence would sell themselves, which looking back on it now with decades more experience, they wouldn't have. I admit my point on the 10% was a bit exaggerated (it was probably about 25-30% at HP) but I was trying to point out that no matter how good the product, if you don't market it well it will usually fail, whereas you can make a poor product succeed with good marketing, as our main two competitors in that field did :-)

Note that marketing isn't all about advertising. Tesla for example spent huge amounts on promoting the early product, providing free long term loan vehicles to hundreds of journalists, including the Top Gear team of course who I believe destroyed two of them. Documentation for dealers to repair the cars also comes under marketing costs, as does the cost of selling and warranty. Tesla is probably unique in not entering motorsport, but every other car company each spends hundreds of millions on that every year as it combines advertising with improving the product and motivating the workforce.

It depends on the company, but recall costs are also often charged to that budget, which for Tesla has been huge, possibly exceeding the cost of R&D. Regulatory fines are another cost which have grown in recent years so you cannot lose under 'miscellaneous' costs any more. So yes, whilst there will be exceptions which for example rely on free advertising through social media, for most successful products marketing costs usually far exceed engineering costs.
Always interested in innovative audio startups needing help and investment. Look for InPoSe Ltd or Future Horizons on LinkedIn to find me (same avatar photograph)

User avatar
lurk101
Posts: 1920
Joined: Mon Jan 27, 2020 2:35 pm
Location: Cumming, GA (US)

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Sep 25, 2022 2:18 pm

MikeDB wrote:
Sun Sep 25, 2022 10:33 am
So yes, whilst there will be exceptions which for example rely on free advertising through social media, for most successful products marketing costs usually far exceed engineering costs.
Well yes, that applies to many product categories. At one point though we had about 150 engineers, with a team of 2 in marketing, and 3 in sales. We constantly cursed the marketing guys for their insane demands and sometimes questionable ethics, but those teams put in long hours and were an essential part of our success as a chip supplier. Over time, despite my initial misgivings, I developed a healthy respect for their work. As far as I recall though engineering was always the dominant cost as I suspect it was for ARM in the beginning. It depends what you're selling.
The Linux Foundation is like loggers who claim to speak for the trees.

timrowledge
Posts: 1422
Joined: Mon Oct 29, 2012 8:12 pm
Location: Vancouver Island

Re: Ars Technica article about the birth of the ARM CPU's.

Wed Sep 28, 2022 2:22 am

“If you look back at it, the first ARM processor wasn't that good”
ORLY? It beat the hell out of intel’s contemporary cpus, at least for my work, which was/still is Smalltalk. I had an ARM1 machine (one of the hand built prototype units - still have it, although upgraded to ARM3 and split clock to make it a 12MHz turbo machine) and it was smokin’ fast for the era.

The StrongARM cpu card for my RISC PC ran Smalltalk faster than its contemporary x86 competition as well. And in fact faster than most contemporary Power PC machines.
Making Smalltalk on ARM since 1986; making your Scratch better since 2012

User avatar
lurk101
Posts: 1920
Joined: Mon Jan 27, 2020 2:35 pm
Location: Cumming, GA (US)

Re: Ars Technica article about the birth of the ARM CPU's.

Thu Sep 29, 2022 9:52 pm

timrowledge wrote:
Wed Sep 28, 2022 2:22 am
“If you look back at it, the first ARM processor wasn't that good”
ORLY? It beat the hell out of intel’s contemporary cpus, at least for my work, which was/still is Smalltalk. I had an ARM1 machine (one of the hand built prototype units - still have it, although upgraded to ARM3 and split clock to make it a 12MHz turbo machine) and it was smokin’ fast for the era.

The StrongARM cpu card for my RISC PC ran Smalltalk faster than its contemporary x86 competition as well. And in fact faster than most contemporary Power PC machines.
A more appropriate comparison would be with other RISC based systems of the era. I don't know of PowerPC machines of the same vintage as the ARM2. The SPARCstation introduced in 1989 would be a close candidate for comparison as the PowerPC wasn't introduced till 1992. Throughout the 1990s and 2000s the MIPS core was a a bigger player in embedded than ARM was.
The Linux Foundation is like loggers who claim to speak for the trees.

pidd
Posts: 4156
Joined: Fri May 29, 2020 8:29 pm
Location: Wirral, UK

Re: Ars Technica article about the birth of the ARM CPU's.

Thu Sep 29, 2022 11:25 pm

SPARC;s were amazing performers, we had a bunch of ICL DRS6000's (also SPARC powered) and I needed another dedicated server for my project, I was expecting something the size of the DRS's to turn up and a Sun SPARCstation rolled up which was a similar size to normal PC. I thought it was never going to cope with the load, far from it. It was accessed by PC'a through XWindows with a lot of users on a heavy database system, I never came across any slow-downs.

Threw out my Sun Solaris (~= Unix SVR4) admin manuals a couple of days ago, I can't believe that Solaris 10 is still supported after all this time.

I've seen a few old SPARC's going cheap in recent times, I've always been tempted but would they perform as well as I remembered?

ejolson
Posts: 10219
Joined: Tue Mar 18, 2014 11:47 am

Re: Ars Technica article about the birth of the ARM CPU's.

Fri Sep 30, 2022 12:13 am

lurk101 wrote:
Thu Sep 29, 2022 9:52 pm
timrowledge wrote:
Wed Sep 28, 2022 2:22 am
“If you look back at it, the first ARM processor wasn't that good”
ORLY? It beat the hell out of intel’s contemporary cpus, at least for my work, which was/still is Smalltalk. I had an ARM1 machine (one of the hand built prototype units - still have it, although upgraded to ARM3 and split clock to make it a 12MHz turbo machine) and it was smokin’ fast for the era.

The StrongARM cpu card for my RISC PC ran Smalltalk faster than its contemporary x86 competition as well. And in fact faster than most contemporary Power PC machines.
A more appropriate comparison would be with other RISC based systems of the era. I don't know of PowerPC machines of the same vintage as the ARM2. The SPARCstation introduced in 1989 would be a close candidate for comparison as the PowerPC wasn't introduced till 1992. Throughout the 1990s and 2000s the MIPS core was a a bigger player in embedded than ARM was.
IBM first had the 801-based ROMP marketed as the RT PC. As summarized in a recent article for a series entitled Computer Chronicles Revisited
S.M. Oliva wrote: As David Patterson noted during the broadcast, IBM began work on what he would later describe as RISC during the 1970s. More precisely, Ferguson and Morris credited IBM’s early RISC work to John Cocke of IBM’s Yorktown Heights (New York) Research Center. IBM kept Cocke’s work secret for most of the 1970s. At the same time, Patterson conducted his own research into RISC concepts at Berkeley–aided by some graduate students who previously interned at IBM–while Dr. John Hennessy led a similar effort at Stanford University. Hennessy later co-founded MIPS Computer Systems, while Hennessy’s work led to the development of the Sun Microsystems SPARCsystem. MIPS and Sun would effectively dominate the early RISC-based CPU market in the United States. (Patterson and Hennessy later shared the 2017 A.M. Turing Award from the Association for Computing Machinery for their work in developing RISC.)
https://smoliva.blog/post/computer-chro ... ibm-rt-pc/

Strangely enough, I spent quite a bit of time with AIX on the IBM RT. I think the CPU design was good enough, but they paired it with an astonishingly slow floating point accelerator.

ejolson
Posts: 10219
Joined: Tue Mar 18, 2014 11:47 am

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Oct 02, 2022 1:01 am

ejolson wrote:
Fri Sep 30, 2022 12:13 am
lurk101 wrote:
Thu Sep 29, 2022 9:52 pm
timrowledge wrote:
Wed Sep 28, 2022 2:22 am
“If you look back at it, the first ARM processor wasn't that good”
ORLY? It beat the hell out of intel’s contemporary cpus, at least for my work, which was/still is Smalltalk. I had an ARM1 machine (one of the hand built prototype units - still have it, although upgraded to ARM3 and split clock to make it a 12MHz turbo machine) and it was smokin’ fast for the era.

The StrongARM cpu card for my RISC PC ran Smalltalk faster than its contemporary x86 competition as well. And in fact faster than most contemporary Power PC machines.
A more appropriate comparison would be with other RISC based systems of the era. I don't know of PowerPC machines of the same vintage as the ARM2. The SPARCstation introduced in 1989 would be a close candidate for comparison as the PowerPC wasn't introduced till 1992. Throughout the 1990s and 2000s the MIPS core was a a bigger player in embedded than ARM was.
IBM first had the 801-based ROMP marketed as the RT PC.
After looking I can’t seem to find an IBM RT PC emulator. Given this was the first commercially produced RISC system, I find it historically odd the software (particularly AOS for which many had the BSD-derived source) has not been preserved through hardware emulation.

I’d sure like a ROMP on my Raspberry Pi.

timrowledge
Posts: 1422
Joined: Mon Oct 29, 2012 8:12 pm
Location: Vancouver Island

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Oct 02, 2022 3:28 am

lurk101 wrote:
Thu Sep 29, 2022 9:52 pm
. I don't know of PowerPC machines of the same vintage as the ARM2.
Nor do I but since I was referring to StrongARM that points at 1996, so consider PowerPC Macs of that era. I had several (both ppc macs and StrongARM RiscPCs).
And indeed I had a SUN 4 for a while, as well as several later Sparc machines.
Making Smalltalk on ARM since 1986; making your Scratch better since 2012

timrowledge
Posts: 1422
Joined: Mon Oct 29, 2012 8:12 pm
Location: Vancouver Island

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Oct 02, 2022 3:32 am

ejolson wrote:
Fri Sep 30, 2022 12:13 am
IBM first had the 801-based ROMP marketed as the RT PC. As summarized in a recent article for a series entitled Computer Chronicles Revisited
S.M. Oliva wrote: As David Patterson noted during the broadcast, IBM began work on what he would later describe as RISC during the 1970s. More precisely, Ferguson and Morris credited IBM’s early RISC work to John Cocke of IBM’s Yorktown Heights (New York) Research Center. IBM kept Cocke’s work secret for most of the 1970s. At the same time, Patterson conducted his own research into RISC concepts at Berkeley–aided by some graduate students who previously interned at IBM–while Dr. John Hennessy led a similar effort at Stanford University. Hennessy later co-founded MIPS Computer Systems, while Hennessy’s work led to the development of the Sun Microsystems SPARCsystem. MIPS and Sun would effectively dominate the early RISC-based CPU market in the United States. (Patterson and Hennessy later shared the 2017 A.M. Turing Award from the Association for Computing Machinery for their work in developing RISC.).
I had one of the prototype 801 boxes when I was an IBM research fellow- worked a little with John Cocke at Yorktown, rather more with the late great David N Smith.
Making Smalltalk on ARM since 1986; making your Scratch better since 2012

n2kra
Posts: 27
Joined: Mon Apr 29, 2013 11:57 am

Re: Ars Technica article about the birth of the ARM CPU's.

Sun Oct 02, 2022 2:58 pm

timrowledge wrote:
Sun Oct 02, 2022 3:32 am
ejolson wrote:
Fri Sep 30, 2022 12:13 am
IBM first had the 801-based ROMP marketed as the RT PC. As summarized in a recent article for a series entitled Computer Chronicles Revisited
S.M. Oliva wrote: As David Patterson noted during the broadcast, IBM began work on what he would later describe as RISC during the 1970s. More precisely, Ferguson and Morris credited IBM’s early RISC work to John Cocke of IBM’s Yorktown Heights (New York) Research Center. IBM kept Cocke’s work secret for most of the 1970s. At the same time, Patterson conducted his own research into RISC concepts at Berkeley–aided by some graduate students who previously interned at IBM–while Dr. John Hennessy led a similar effort at Stanford University. Hennessy later co-founded MIPS Computer Systems, while Hennessy’s work led to the development of the Sun Microsystems SPARCsystem. MIPS and Sun would effectively dominate the early RISC-based CPU market in the United States. (Patterson and Hennessy later shared the 2017 A.M. Turing Award from the Association for Computing Machinery for their work in developing RISC.).
I had one of the prototype 801 boxes when I was an IBM research fellow- worked a little with John Cocke at Yorktown, rather more with the late great David N Smith.
Was DIVMOD faster, useful in the numeric tower?
barrel shift per instruction.(like Sparc 32 fixnum?)
Neon/AVX acceleration (encryption, media)
Pi 400/4/3B+ - Ham, Tech Net, Linux / Mac / PC, Programmer (for IF author).

Return to “Off topic discussion”