Alasir Enterprises
Main Page >  Articles >  Alpha: The History in Facts and Comments  

 
Main Page
 
 
Reviews
 
Articles
 
Software
 
Reference
 
Motley
 
 
About Us
 
 
Alpha: The History in Facts and Comments
Alpha Powered Dig my grave both long and narrow
Make my coffin neat and strong 

(from an old American song)

Paul V. Bolotoff
 
Release date: 14th of April 2005
Last modify date: 6th of February 2011

Contents:
in Russian

 
The Collapse of DEC

On the 26th of January 1998, a news flashed all over the computer world that struggling financially DEC was purchased by Compaq Computer Corporation and the deal was about to be approved by the upcoming shareholders' meetings of both industry giants. Shareholders of DEC ratified the agreement on the 2nd of February 1998. The amount of sale was 9.6 mlrd. USD, much higher if compared to the company's market capitalisation prior to the acquisition of about 7 mlrd. USD. The process of integrating DEC's functional units into Compaq's business structure had been completed about half a year later with the legal end of DEC, when its shares were taken off the New York Stock Exchange on the 11th of June 1998. To mention, negotiations between DEC and Compaq started in 1995 but finished unsuccessfully in 1996, because top management of DEC held a position insisting on a merger, not an acquisition. Nevertheless, here comes a question: how could it happen that a huge company (in figures of 1989: almost 130 thousand of personnel, gross revenue of about 14 mlrd. USD per year, i.e. the second largest company in the industry after IBM) which held a very high R&D potential and significant manufacturing facilities, was forced to sell itself to a large computer building company from Texas? There was no single answer to this question, though reasons mentioned were numerous and various.
 
A long time ago, Kenneth Olsen, who was a founder, president and CEO of DEC until almost the end, said that well engineered products would sell themselves. Thus, they have no need in any advertising campaigns or other instruments of market promotion. He also mentioned that there is no reason anyone would want a computer at home. Perhaps, these thoughts were correct in those "old good times" when computer equipment was manufactured in limited quantities by professionals and for professionals, thus cost a hefty amount per unit. However, they weren't appropriate somewhere close to the end of the 20th century when computer equipment was sold in million units per year, and a very regular computer could be taken together using a screwdriver and parts from the nearest computer shop for an hour of free time maximum. Besides, it would cost over 10 times less than a wardrobe-like one from those "old good times" mentioned previously. Finally, nothing should prevent you from ordering a whole working box right from that shop with a free delivery. Considering that such a regular machine would be purchased most likely not by a professional manager realising clearly what TCO (Total Cost of Ownership) stands for, but by an aunt Marge or a young prankster Johnny making no difference between transistor and resistor, so such customers should be motivated definitely not by engineering advantages of a potential purchase.
 
At the very beginning of the Alpha architecture's way, DEC's top management made a great strategical mistake. First prototypes of EV4 were presented on a computer conference in February of 1991. There were engineers of Apple Computer admitted among others, and they were looking for a new processor architecture to power their company's future computers. Of course, they were impressed by advantages of EV4. John Sculley, Apple's CEO of those days, met with Kenneth Olsen in June of the same year and offered him to use the new processor of DEC in future Macs. Olsen refused the offer motivating that EV4 was not ready for the market, besides the VAX architecture hadn't reached its end-of-life yet. Several months later, rumours said that new Macs would be powered by PowerPC processors designed and manufactured by the alliance of Apple, IBM and Motorola. So, DEC had lost an excellent opportunity to achieve good long term sales of Alpha hardware and to promote the architecture on the desktop market, but gained a competitor instead. William Demmer, a former vice-president of VAX and Alpha divisions who resigned in 1995, said in his interview to the Business Week published on the 28th of April 1997: "Ken did not want the company's future to run on Alpha."
 
DEC manufactured Alpha processors as well as accompanying system logic sets and numerous peripherals at the factory of its own in Hudson (Massachusetts, the USA). It designed and produced OEM available mainboards for desktops and workstations called "Evaluation Board" or "AlphaPC". Their assortment was rather limited though. Neither of them supported multiprocessing, though all DEC Alpha powered servers except of entry level models were multiprocessors. Nevertheless, all mainboards were engineered very well, though expensive very much like Alpha processors. Their layout schemes were available for public access, so several companies (Aspen, Polywell, Enorex asf.) manufactured fully qualified clones. The only company to develop and produce proprietary designs was DeskStation. In general, it could be stated for sure that DEC considered a priority to produce workstations and servers of its own, but didn't bother itself to fill the market with their basic components. Of course, this approach offered more income in short perspective, but there was no chance to establish a strong presence on the desktop or workstation market this way.
 
Despite all attempts taken, DEC failed to make prices (considering processors, system logic sets and mainboards first) affordable to most potential customers. For example, 266MHz and 300MHz EV5 were offered in the beginning of 1995 for 2052 and 2937 USD respectively in lots of 1000 units — both prices were enormous even if to take into account average estimated manufacturing costs of 430 USD per unit. Considering price per one "parrot" of SPECint92, EV5 cost about 2 times higher than competitive RISC designs. At the same time, a standard system logic set for it, DEC Alcor, was offered much cheaper — 295 USD each in lots of 5000 units. Nevertheless, the only Alcor based mainboard from DEC, EB164 with 1Mb of B-cache, bundled with a processor and 16Mb of operating memory (by the way, that wasn't enough to run many applications on Alpha even those days), carried a list price of about 7500 USD.
 
Although Alpha was declared an open architecture right from the start, there was no consortium to develop it. All R&D actions were handled by DEC itself in cooperation with Mitsubishi sometimes. In fact, though the architecture was free de jure, most important hardware designs of it were pretty much closed de facto, so had to be paid licenced if could be at all. Of course, it wasn't that thing helping to promote the architecture. For instance, if they kept EV4 and EV5 designs proprietary but let LCA4 go free without any licence fees or other obligations implied, it would be a strong move shaking the desktop market well. Soon after the introduction of EV4, DEC offered to licence manufacturing rights to Intel, Motorola, NEC and Texas Instruments. However, all these companies were involved in different projects of their own and didn't want to play the second fiddle at this party, so they refused. Perhaps the conditions could be also unacceptable or something else.
 
After all, even the fastest computer without an operating system and accompanying software is just an expensive source of noise and an environmental heater. DEC targeted its Alpha hardware for Windows NT, Digital UNIX and OpenVMS following this priority order exactly. Could be not bad, but...
 
Windows NT was an operating system designed for users when right out of the box, not for programmers since there were no software development tools supplied. Hence, it was dependent heavily upon precompiled applications, commercial notably. In fact, numbers of Alpha- and i386-ready software titles for Windows NT were different by a few times. Although there was FX!32, an excellent emulator and translator of i386 code to Alpha released by Anton Chernoff's team in 1996. While being a useful solution itself, couldn't help with performance decrease of 40% on average when compared to the same source code compiled natively. Next, there were drivers and FX!32 was absolutely of no help in this area. Considering a fact that very few hardware manufacturers honoured the Alpha architecture enough to release any of them, users had to rely mostly upon Microsoft and DEC. Finally, Windows NT (3.51 as well as 4.0) was a 32-bit OS regardless of running on the 64-bit Alpha hardware, thus was unable of utilising it to the full extent. However, all these issues didn't prevent DEC from promoting its Alpha systems with a slogan "Born to run Windows NT". In brief, such an OS shouldn't be positioned as the primary for the Alpha architecture, though having it available as an option was a big plus on the workstation market.
 
OpenVMS OpenVMS for Alpha originated from the 32-bit VAX/VMS operating system developed since 1976 for DEC VAX machines. In 1990, DEC programmers started on a porting project of VAX/VMS 5.4-2 to Alpha, and this task had been finished by November of 1992 with the release of OpenVMS AXP 1.0 followed by OpenVMS VAX 6.0 in June of 1993. There was some mess with the numbering because the next release of OpenVMS was 6.1 both for Alpha and VAX in April-May of 1994. The next major release — 7.0 — took place in December of 1995. Among numerous improvements, it allowed finally to use the 64-bit virtual addressing on Alpha. Compaq started to port OpenVMS from Alpha to Itanium in 2001, and this project had been completed mostly by the 8.0 release in June of 2003. The first unified release of OpenVMS for the Alpha and Itanium architectures was 8.2 (February of 2005). In August of 2000, Compaq announced the end-of-life for VAX, so the development of OpenVMS for VAX did not extend past 7.3, though standard support was still active through Hewlett-Packard as of 2011. OpenVMS 8.4 for Alpha and Itanium (release date: June of 2010) is the current version developed, promoted and supported by Hewlett-Packard.
 
Digital UNIX has a different background. The first UNIX workstations by DEC were powered by Ultrix, a proprietary 4.2BSD derivative with some AT&T System V features supported. It was released initially in June of 1984 for the VAX and PDP-11 architectures, later ported to MIPS (1988) to serve as a primary OS for the company's first RISC workstations: DECstation 2100, DECstation 3100 and DECstation 5000. These were in production until 1994 when were obsoleted by Alpha powered AlphaStations. Some Ultrix code was a property of AT&T, so DEC was bound by a respective licence agreement. Those were times of UNIX related legal wars, so DEC together with IBM, Hewlett-Packard and other top brands established the Open Software Foundation (OSF) in 1988 to oppose AT&T and Sun Microsystems. It released a UNIX-like operating system called OSF/1 in December of 1991 which was free of any AT&T code. DEC offered it for almost the entire line of its Alpha workstations and servers under the name OSF/1 AXP. Although some models were promoted for Windows NT only as they lacked firmware necessary to run OpenVMS or any UNIX derivative. OSF stopped the development of this OS in 1994, so DEC continued on its own and renamed it to Digital UNIX with the 3.2 release (1995). The name was changed once again by Compaq to Tru64 UNIX with the 4.0E release (April of 1999). The current version 5.1B-6 (release date: October of 2010) is developed, promoted and supported by Hewlett-Packard.
 
So, here is the summary. OpenVMS and Digital UNIX are reliable and scalable commercial operating systems, but they didn't obtain any really significant popularity back in the time because of high prices. For instance, a single 2-user licence of Digital UNIX 4.0 cost 795 USD in 1996. Of course, their source code was closed. Although there were other drawbacks present such as even more limited hardware base supported when compared to Windows NT, if either of these OS was given freedom together with DEC's excellent development tools, it could increase the Alpha architecture's market share strongly.
Digital UNIX
NetBSD DEC didn't support free open-source operating systems, though the very first of them, NetBSD, was ported to Alpha in 1995, followed by Linux, OpenBSD and FreeBSD. It was strange somewhat because these OS were (and still are) popular very much in the Alpha world. In addition, their high potential was apparent even those days and was increasing constantly. Besides, these OS featured no worse performance than commercial Digital UNIX or OpenVMS and hardware support comparable to Windows NT (much better nowadays) as well as many other benefits you may expect from open source software.
 
FreeBSD has abandoned support for Alpha since the 7.0 release (February of 2008). NetBSD, OpenBSD and Linux continue to support the architecture.
OpenBSD
Linux FreeBSD

This list of DEC's great mistakes could be expanded even more. However, the other ones are either not important that much as those mentioned above or not related directly to the Alpha architecture. Anyway, the following final conclusion could be derived from the author's point of view: DEC had done a real number of efforts to make as much money as possible with the Alpha architecture, but had done almost no efforts to help the architecture itself.
 
The board of directors which was motivated by numerous company's failures during the late 1980's and early 1990's, suspended Olsen from managing the corporation in June of 1992 and appointed Robert Palmer instead. He did a hard try to reorganise the company's business in 1994 by turning the existing matrix model (where departments different functionally cooperated to make a decision) into a traditional vertical (where authorities and responsibilities were defined clearly from the very top to the very bottom of an enterprise). From 1991 to 1994, the company's net losses figured into over 4 mlrd. USD including 2.16 mlrd. just from July of 1993 to June of 1994, though those in turn included 1.2 mlrd. spent on restructurisation which involved cutting about 20 thousand jobs. So, the total number of personnel was reduced to 85 thousand. According to Palmer's programme, the company had to get rid of many divisions considered non-priority, so the global sale began.
 
In July of 1994, the Storage Business Unit manufacturing disk and tape drives was sold to Quantum for 348 mln. USD. Well, this division was out of favour due to a fiasco of the company's first thin film hard drives (DEC RA90 and RA92 — 1.22GB/3600rpm and 1.51GB/3400rpm respectively, DSA/SDI interface). They were expected to hit the market by the end of 1988, but it didn't happen until summer of 1990 due to design flaws. As a result, DEC lost their leadership in the HDD business to Imprimis, a subsidiary of Control Data Corp. (CDC), which announced their thin film Wren VII 1.05GB/3600rpm 5.25" full height SCSI drive in November of 1988 and actually delivered it to the market next year as Seagate ST41200N/ST41200ND (Imprimis was sold to Seagate in June of 1989 for about 450 mln. USD in cash and securities). Therefore, Seagate appeared to be the first storage company to break throgh a 1GB barrier for HDDs. Maxtor to follow, and DEC to manage losses. About Quantum, it did have a quite successful HDD business back in the time, but their products had never reached high end before and were very popular on the lowest end (Quantum Bigfoot series HDDs produced since 1996 in 5.25" form factor and 3600rpm rotation speed were an excellent example). They wanted to expand their operations to the profitable SCSI market, and DEC had a lot of experience in this area. DEC storage products of the StorageWorks family were based upon RZ series SCSI HDDs. Some of them were manufactured by Micropolis, Conner, Quantum and others for DEC, though some models were proprietary designs engineered by a team in Shrewsbury (Massachusetts, the USA). They also designed and manufactured successful DLTs (digital linear tape drives) since 1984 which were another acquisition objective for Quantum. In brief, the former DEC technologies and personnel allowed Quantum to establish its Atlas family of 7200rpm SCSI drives, later expanded to 10K and 15K rpm. This product family continued to be developed after the 50%/50% merger of the Quantum HDD business with the whole Maxtor company in October of 2000, and was phased out soon after the 16%/84% acquisition of Maxtor by industry leader Seagate in December of 2005 because of Seagate's own Cheetah product family of high speed SCSI HDDs.
 
In August of 1994, the Database Software Unit was sold to Oracle for 108 mln. USD in cash. The deal included DEC Rdb relational data base management system (RDBMS) as well as CDD/Repository and DBA WorkCenter software. Although DEC RMS (Record Management Services) data base software was excluded from the deal because it was integrated deeply into OpenVMS and couldn't be removed. DEC supported Rdb 6.0 on OSF/1 AXP, OpenVMS (VAX and Alpha), Windows NT (i386 and Alpha). With the release of Rdb 7.0, Oracle had discontinued support for Rdb on all those operating systems except OpenVMS, and phased out its VAX branch in favour of Itanium later.
 
DEC had established an alliance with Olivetti, an Italian computer vendor, in 1992 to promote products mutually. DEC had also purchased a 7.8% share in Olivetti for 287 mln. USD. On the 24th of August 1994, DEC redeemed this share for 150 mln. USD. That was almost half the price paid 2 years ago. Not the best investment indeed. Although both companies continued their partnership until 1997 when Olivetti decided to shut down its computer business.
 
In November of 1997, the Network Product Business Unit was sold to Cabletron for about 430 mln. USD in cash, securities and product credits. In addition to tangible assets, Cabletron did acquire about 250 network related patents and 900 employees not including those 200 people who were lucky enough to have their jobs kept at DEC. The amount of those product credits was undisclosed, so it was difficult to estimate the real sale price. By the way, Cabletron laid off about 600 people after shutting down two manufacturing facilities next month, and lost about one half of its market capitalisation for that single month since the deal with DEC. In fact, Cabletron wasn't really interested in those DEC network products, or technologies, or even employees. They wanted the sales network. However, it didn't work out. The company jumped from one trouble to another over the following years. Finally, remains of the former DEC network division with 130 employees were sold to Gores Technology Group in May of 2000.
 
The fall of DEC was loud enough. It sued Intel in May of 1997 accusing in infringements upon 10 patents issued for the Alpha architecture while designing Pentium, Pentium Pro and Pentium II processors. Intel started a lawsuit against DEC in September of 1997 claiming its 14 patents to be dishonoured while designing Alpha processors. The peace was reached finally on the 27th of October 1997 when both companies took their complaints back. DEC licenced to Intel manufacturing rights for all its hardware available except of Alpha property, also agreed to support the future IA-64 architecture. In exchange, Intel purchased from DEC the factory in Hudson together with designing centres in Jerusalem (Israel) and Austin (Texas, the USA) for 625 mln. USD, also agreed to manufacture Alpha processors for DEC in the future. Additionally, an agreement was signed to cross-licence patents for 10 years. The deal was completed on the 18th of May 1998. By that time, Compaq had started to incorporate DEC's primary divisions with about 38 thousand of personnel. Many of them were laid off in the very near future though. By the way, Compaq employed about 32 thousand of personnel on its own prior to the acquisition.
 
It needs to mention that not so long before the end of DEC and soon after many talented engineers who actually created that DEC's realm left for other employers. Derrick Meyer and James Keller went to AMD to design K7 and K8 respectively. Daniel Leibholz was hired by Sun to create UltraSPARC V. Richard Sites, one of the primary Alpha architects, also abandoned the new owner. Intel appeared to be lucky even less. The StrongARM architecture which was purchased from DEC together with the factory and designing centres seemed to be at a dead end because none of those chief architects who developed StrongARM-110 previously (Daniel Dobberpuhl, Richard Witek, Gregory Hoeppner and Liam Madden) decided to join the Chipzilla. Even more, Witek's team which worked in Austin towards the next StrongARM core resigned completely. So, Intel had to relocate engineers who developed i960 previously to work on this project literally from the scratch.
 
<< Previous page Next page >>

Copyright (c) Paul V. Bolotoff, 2005-11. All rights reserved.
A full or partial reprint without a permission received from the author is prohibited.
 
Designed and maintained by Alasir Enterprises, 1999-2011
rhett from alasir.com, walter from alasir.com