[Info-vax] Sunway TaihuLight is the fastest supercomputer in the world

[Kerry Main]

> -----Original Message-----
> From: Info-vax [mailto:info-vax-bounces at rbnsn.com] On Behalf Of
> johnwallace4--- via Info-vax
> Sent: 25-Jun-16 5:17 AM
> To: info-vax at rbnsn.com
> Cc: johnwallace4 at yahoo.co.uk
> Subject: Re: [Info-vax] Sunway TaihuLight is the fastest supercomputer
> in the world
> 
> On Wednesday, 22 June 2016 22:17:48 UTC+1, lawren... at gmail.com
> wrote:
> > On Thursday, June 23, 2016 at 12:14:49 AM UTC+12,
> johnwa... at yahoo.co.uk wrote:
> > > DEC did let others make Alpha. At least Samsung and Mitsubishi had
> > > the kind of Alpha licences that let them design and make their own
> > > Alpha chips. Samsung used theirs to design chips and boards of
> > > their own, not just clones, which carried the AlphaPowered logo,
> > > before times changed and suddenly Alpha wasn't a promising place
> > > to be any more.
> >
> > Why did that happen? Was Alpha too expensive? Did DEC’s lawsuit
> against Intel scare the customers away?
> 
> 
> There was indeed a lawsuit, a public settlement, and some
> backroom deals.
> 
> At the time, DEC HQ were persuaded that Intel (either with
> x86 or IA64) were more important to DEC than Alpha would be.
> Not least because Intel had huge piles of cash and therefore
> anything they did would be successful. DEC, on the other hand,
> wasn't flush for cash.
> 
> The overall cost of devloping a chip (and any associated
> manufacturing facilities) is largely independent of the volume
> sold.
> 
> If there's no cash pile to support development, the next
> likely source of money is the chip's customers. This led to
> what was called the "Alpha tax"; regardless of the actual chip
> manufacturing costs,someone's still got to pay for development.
> 
> Intel, in contrast, had both a huge cash pile and an ongoing
> high-margin high end chip business (Xeon etc).
> 
> Volume's a part of that picture too. But not the only important
> part. Another part was that DEC HQ appeared to think that they
> could either be Intel's best friend (without Alpha), or they
> could go their own way (with Alpha).
> 
> This was at a time when some people in DEC HQ still thought
> there was useful profit to be made from designing and building
> robust well engineered x86 PCs for the business market.
> Ignoring what would turn out to be a high volume low margin
> barely profitable business was not thought to be an option. Oh
> well.
> 
> For a while, DEC's later low end Alpha designs shared a great
> deal of hardware with the equivalent x86 DEC US PCs. For example,
> the Personal Workstation series had Alpha members as well as
> x86 members, all based around NLX form factor motherboards with
> processor-specific daughter cards. The Alphastation 400 had
> tried the same common box+motherboard, different daughtercard,
> principle previously.
> 
> They did OK in some specific markets in some specific geographies,
> where they were targeted at markets where they had advantages.
> But against a commodity x86 NT PC from Gateway2000 etc they stood
> little chance.
> 
> But the backroom deal associated with the court settlement ensured
> that regardless of the technology, Alpha had no real future because
> IA64 was going to take over the world of "industry standard 64bit".
> 
> And in due course that's where AMD64 came in (as mentioned by IanD
> and did what Intel had been saying was impossible - a 64bit
> x86-compatible chip family. And the rest is history.
> 

What John stated + propaganda by Intel showing DEC what they were 
expecting to achieve in 3 years on X86 really scared DEC mgmt. 

Btw, Intel to this day not reached the CPU Ghz numbers they were 
pitching back then.

Let's not forget where a significant boost in X86 performance came from
(Alpha inside) and that was the reason for the lawsuit which Intel lost.

Bottom line, like a card game, Intel bluffed with what they had and DEC 
blinked. 

Imho - At some point in the late 80's timeframe, DEC mgmt. did not know 
what kind of company they wanted to be when they grew up. 

Hence, they sold key SW products, then sold key HW prod's, then the board 
and C levels threw their hands up and decided to line up at the trough for
big bonuses by selling the company.

At the time of sale to Compaq, DEC had $3B in the bank which was approx.
30% of what the sale was for ($10B). No doubt the fire sales leading up to
the final sale were meant to sweeten the selloff.

Ahh well, water under the bridge .. this type of corporate shenanigans
Is common place today.

For another view of DEC mistakes-
http://www.sigcis.org/files/Goodwin_paper.pdf 
" In 1987 Digital Equipment Corporation (DEC) was the number two 
computer manufacturer in the world with its founder being named the 
“‘most successful entrepreneur in the history of American business” by 
Fortune magazine. This paper looks at the later history of Digital Equipment 
Corporation and asks how an organization that was so successful in 1988 
could sink to become a takeover target for a PC hardware company ten 
years later."

Regards,

Kerry Main
Kerry dot main at starkgaming dot com