Monday, July 7, 2014

More about 'The Machine'

Just saw Bill Harrell's description for Digital Trends of HP's "Machine"     See it at:
http://www.digitaltrends.com/computing/what-is-the-machine-hps-new-super-computer/#!94DVM

Might be worth considering how this timeline compares with Birnbaum's Spectrum from the '80's.

We hired Joel in January 81, and by 83 he had some nice prototypes running at HP Labs.  By fall '84, Birnbaum and Doyle moved to Cupertino to revamp the computer group leadership, a group which had been resistant to RISC ideas (and to HP Labs in general).

In May 1985, Bill Worley among others outlined HP/PA (HP Precision Architecture) in the HP Journal.  In February 1986, HP had a major press conference release to announce the machines.  In a stern voice at the end of the press meeting, Packard dolefully pointed out that the machine was not ready, pissing off everyone else on the dias.

In November 1986, the first machine shipped, not exactly a big hit, but at least 'on the street'

Between 1987 and 1989, the big news was that the new migration center was able to make the novel HP RISC architecture handle substitution into DEC and IBM bastions--not yet outperforming them, but at least able to offer an alternative, usually much cheaper and in places more reliable.

In Febrary 1991--yes, just ten years after starting--the Snakes project--the HP 700-- led by Denny Georg, was announced and delivered.  It was the first RISC machine with industry-leading performance.

Now, granted, things happen quicker these days, and designers and managers are smarter, right?

But, the HP plans for "the Machine" sound just a bit optimistic.  Hope they're right!

6 comments:

Me said...

You write:

We hired Joel in January 81, and by 83 he had some nice prototypes running at HP Labs. By fall '84, Birnbaum and Doyle moved to Cupertino to revamp the computer group leadership, a group which had been resistant to RISC ideas (and to HP Labs in general).

The cancellation of the Vision Computer Family (VCF) left a really unholy mess. VCF was basically the byproduct of internecine warfare between GSD (HP3000) and DSD (HP1000) and TSD (HP9000). It was designed to make everyone happy and, in reality, probably wasn't going to make anyone happy.

Indigo didn't make anyone happy either (the raw performance was there but the software really bogged things down -- especially MPE-XL...MPE-XL was so slow at the beginning that you could count the time between pressing a key on the keyboard and the appearance of the character on the screen in seconds. As I recall, the holdup on PA-RISC was all due to MPE-XL...HP-UX had been ready for years prior to MPE-XL.

Me said...
This comment has been removed by the author.
Me said...


You write:

In Febrary 1991--yes, just ten years after starting--the Snakes project--the HP 700-- led by Denny Georg, was announced and delivered. It was the first RISC machine with industry-leading performance.

When I worked at SID, I had one of these on my desk — a 755 I think (the model numbers were 710, 725, 735 and 755). It was probably the best experience I ever had with an HP workstation! It was wicked fast and I had done a lot of customization on HP VUE and XEmacs to make it match my work style. I still use some of that stuff today (the parts that run on Linux, anyway). When I left SID, I tried to negotiate taking it with me but, in true HP fashion, no soap…

When I got back to Cupertino in 1999, I had a 735 but I had to use NeXTStep on it (I was working on a doomed integration of NeXTStep Portable Distributed Objects and CORBA… PDO was great, well-defined and easy to work with…it just worked…while CORBA was designed by a committee…specs were cumbersome and ambiguous…had every freakin’ feature in the world no matter how useless and was DOG SLOW…I hated it). Most people don’t realize that NeXTStep was a supported OS on PA-RISC for a time.

After that experience, the division was re-purposed and we all had to start using Windows NT (this was the time of Rick Belluzzo’s “NT is taking over the world” push). Yuck! I never worked on an HP workstation (I think HP got out of the workstation market shortly thereafter) although I junked NT for SuSE Linux as soon as I could.

Me said...

You write:

In Febrary 1991--yes, just ten years after starting--the Snakes project--the HP 700-- led by Denny Georg, was announced and delivered. It was the first RISC machine with industry-leading performance.

When I worked at SID, I had one of these on my desk — a 755 I think (the model numbers were 710, 725, 735 and 755). It was probably the best experience I ever had with an HP workstation! It was wicked fast and I had done a lot of customization on HP VUE and XEmacs to make it match my work style. I still use some of that stuff today (the parts that run on Linux, anyway). When I left SID, I tried to negotiate taking it with me but, in true HP fashion, no soap…

When I got back to Cupertino in 1999, I had a 735 but I had to use NeXTStep on it (I was working on a doomed integration of NeXTStep Portable Distributed Objects and CORBA… PDO was great, well-defined and easy to work with…it just worked…while CORBA was designed by a committee…specs were cumbersome and ambiguous…had every freakin’ feature in the world no matter how useless and was DOG SLOW…I hated it). Most people don’t realize that NeXTStep was a supported OS on PA-RISC for a time.

After that experience, the division was re-purposed and we all had to start using Windows NT (this was the time of Rick Belluzzo’s “NT will rule the world” push). Yuck! I never again had an HP workstation (I think HP got out of the workstation market shortly thereafter) although I junked NT for SuSE Linux as soon as I could.

chuck said...

Another reader sent these thoughts along:

After reading your blog this morning I couldn't resist to add my comments. You may have noted that some pundits like Cringeley panned the HP Machine as largely vaporware. Your timeline today with earlier HP history adds a useful perspective.

What is missing from the current discussion thus far is a realistic technology assessment, i.e. are the two key technologies chosen the right ones? Fink is no Barney Oliver and I have no faith in his technical judgment. One technology, silicon photonics, will likely be important but there are many players way ahead of HP in this technology - what remains of the eviscerated HP Labs effort is pathetic.

The second technology, memristers, has been over hyped for years now by HP. Aside from academics, few in the industry see much future for this. Again the hard probing questions of Barney (and sometimes Bill Hewlett) are missing here - where is the informed technical judgment?

I recall other hyped technologies that were going to change the landscape, but did not. One of these was the Ovonic switch, and Ovshinsky himself came to HP Labs to sell the idea (but of course didn't succeed). We knew the right questions to ask and had a good grasp on where mainstream and competing technologies were headed.

Another one HP Cupertino did buy into was Silicon on Sapphire (SOS), an exceedingly poor technical choice for the time and application.

Hopefully there will be more evaluation about what HP has (and doesn't have) this time!

If you would like to discuss further over coffee or luch next time you're in the Bay Area let me know!

chuck said...

Still another reader input:
Incidentally, these early technology choices were not of a "bet your company" nature, and Cupertino quickly recovered from the SOS fiasco. (Actually, it wasn't 'quickly' but took nearly a decade).

As you know, the scene of that action (HP Cupertino computing group) is being reduced to rubble to provide room for Apple to grow (including expanding their chip design capabilities). The current situation at HP is much closer to a "bet your company" wager!

Ely learned his lesson from SOS experience and later advocated only mainstream technologies for the Computer Group. We were still encouraged to explore new and advanced technologies for the Instrument side, and some of those still live on today!