The Legend of Amdahl [Hardcover]

For over twenty years, Amdahl Corp was a great place to design and build interesting computers.

This book captures little of what made it great, and ignores or glosses over the joys, the problems, and the challenges that were central to the Amdahl experience as it was lived. Instead, the author gives a lifeless, through-the-gauze-lens account that reads like a compilation of Amdahl PR releases and internal newsletters.

The most glaring omission is the complete absence of any mention of Amdahl's ill-fated "Key Computing" venture -- a $350 million boondoggle that sapped the company's resources at the exact instant that Amdahl faced the make-or-break challenge of precipitously-declining price/performance in the mainframe market.

Also missing and missed:

- A recounting of the company's near-death experience with the 580 ("Oslo").

- Any indication that the company might have successfully continued as a competetive mainframe supplier, if it had made the bipolar-to-CMOS technology transition in step with IBM. The 5990M ("Joshua", later "Sona"), an ECL machine that required elaborate power and cooling arrangements, was much more expensive to manufacture than IBM's competing CMOS machine.

- Any hint of any downside to the relationship with Fujitsu.

Finally, I was disappointed by the choice of photographs. Amdahl was above all a *community* of outstanding people, and that community is conspicuous by its absence from these pages. The photos chosen for inclusion lean heavily toward the most stiffly-posed of glossy PR shots -- few former employees will go misty-eyed with nostalgia over these shots, because almost none of the people who made the place what it was appear. Every photograph of professional models standing before the newest box seems to me a wasted opportunity to actually tell the story promised in the title.

A telling detail: in the days when Amdahl Corp actually was a computer-industry legend, the company was known by its all-lower-case logo, often represented as

a m d a h l

The editors chose instead for this book's cover the "new" logo, the one adopted after Fujitsu took over, after everything that a person could love in the organization had died.

There's an Amdahl story that deserves to be told. This book doesn't tell it.

the other Joel

The book is a puff-piece from the perspective of Fujitsu Limited. A rather distorted view of one of the true legends of Silicon Valley.

The ways in which Fujitsu contributed to Amdahl's success are overemphasized yet the ways in which Fujitsu inhibited Amdahl's potential for success ... the opportunities for success not taken at all and the Fujitsu-preferred opportunities for success taken in preference to the Amdahl-preferred opportunities for success ... are not mentioned at all.

Roughly one-half of the book is devoted to the largely unsuccessful Fujitsu initiatives after the decision was made not to proceed with a 64-bit competitor to IBM's z/System, although Amdahl was in a unique position to successfully do so, a decision which doomed Amdahl to the scrap-heap of Silicon Valley companies who "could-a, should-a and would-a," had it not been for the blatant interference from Fujitsu Ltd and Fujitsu America Inc.

The section on communications products mentions the purchase of Tran. The largely unsuccessful integration of Tran into Amdahl, which never produced significant revenue yet it was a constant drain on Amdahl's resources (about $500,000,000 in net losses over its fortunately brief history) is, of course, down-played. The hugely successful Amdahl-Fujitsu joint development of IBM 3705, 3725 and 3745 communications processor competitors, the Amdahl 4705, 4725 and 4745, isn't mentioned at all.

The failure to acquire EMC, although Amdahl had two opportunities to do so, and the abortive attempts at internally developing an EMC-like product isn't mentioned at all, yet the joint Amdahl-Fujitsu development of IBM look-alike direct access storage facilities (as mainframe hard disk drive systems are called) which at several points fueled Amdahl's overall profits as the losses in other areas and the delays of the 5995M and other products mounted, is down-played.

The less said about Amdahl's ill-timed and often mad dash for the so-called Malcolm Baldrige National Quality Award the better. This initiative by now deceased Amdahl president Joe Zemke defocused Amdahl's entire management team, from the bottom to the top, and caused critical delays in Amdahl-designed mainframes, which ultimately forced Amdahl into accepting Fujitsu's not-ready-for-prime-time mainframes until Amdahl's mainframes finally became available.

No, this book is really more about how the third-rate Fujitsu (after the first-rate NEC and the second-rate Hitachi) meddled with Amdahl, both the corporation (which it ultimately doomed), and, yes, the man (who was soon forced out), if the reader will only read between the lines through objective (not rose-colored) lenses.

Both Joel's have it right, IMHO. I knew "the other Joel" well during my time at Amdahl (is it only ex-Amdahl employees that buy this book? :-) ) and agree with his assessement, but would add that there was a passing mention of the 580 experience (see page 64, but they certainly didn't call it "near-death"!), the Key Computing Lab venture, and the problems with Joshua/Toro/Sona.

But not in the detail or accuracy that was appropriate. The problem is that the interviews were with engineers who started with the company early and were in upper management positions when things went sour. They weren't in the trenches fixing the problems, and perhaps weren't willing to make painful observations, e.g. that the decision to continue with ECL instead of designing a CMOS machine was short-sighted. (It actually led to higher revenues in the short term because our ECL machines were faster than IBM's CMOS machines, but ours were much more expensive to manufacture, which led to profitability problems.)

I was in the trenches working on the 5995M, and the observation on page 101 that the 5995M suffered from "some technology problems from Fujitsu" is quite an understatement. Details of the problems and their hard won solutions would have been far more interesting than the bland "I knew we would fix it" statements. And make no mistake, all of these issues were the beginning of the end for a great company. In 2000, when this book was written, there was still a company left, but since then it has split further (Amdahl Software became Softek for example) and going to www.amdahl.com now takes you to www.computers.us.fujitsu.com.

Still, I enjoyed reading the book. It has some shortcomings, but if you were part of that experience, as I was, I think you'll be glad you read the book. Whether or not it's worth a hardcover price is another question.

› Go to Amazon.com to see all 5 reviews  3.2 out of 5 stars