ARCHIVES
The Data Storage Report - April 1996 Volume 11, Issue 4


1996
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1995
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1994
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1993
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1992
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1991
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1990
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1989
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1988
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1987
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1986
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1985
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH
FEBRUARY
JANUARY

1984
DECEMBER
NOVEMBER
OCTOBER
SEPTEMBER
AUGUST
JULY
JUNE
MAY
APRIL
MARCH

 

 

CMI'S IBM CONTRACT, BLESSING OR CURSE

Editor’s note: The following article appeared in the second issue of Data Storage Report in July 1985.

Almost as soon as the IBM PC AT was announced there were news story accounts of troubles in the computer's disk drive. At the same time came revelations that IBM was backordered on higher performance multi-user IBM PC ATs and had a surplus of IBM PC XT's, the previous generation single-user hard disk system. One question Electronic News November 12, 1984 asked was, is the disk drive problem causing the lack of availability?

Press reports layed blame for the IBM PC AT problem on the CMI disk drive. In addition, CMI reported an $8 million loss for the third quarter ending December 31 1984, this at a time when the original IBM contract was expiring and a new one was being negotiated. With all these developments, what are the prospects for CMI? Is it still a company blessed by an IBM contract or one cursed by being at the mercy of one large customer? Will being an IBM supplier strap the company's resources so thin that it is unable to develop a next generation product, a problem that befell International Memories, Inc. (IMI) of Cupertino, California. The company sold off its disk drive operation last year. It sold the half-high 5 1/4-in. product line to CMI.

IBM PC AT drive having trouble

The August, 1984, the disk drive industry learned that Computer Memories, Inc. (CMI), became the sole source to IBM for the IBM PC AT. Computer Systems News, August 20, 1984 made the following observation. "Reliable sources said that between this past April 1 and next March 31, IBM will purchase about $90 million worth of drives from CMI. Average selling price of the drive currently is believed to be about $535 with the price to IBM expected to drop to $500."

In retrospect, there seems to have been two problems. One was IBM underestimated the demand for the AT. A second was the problem in the disk drive subsystem. CMI says that IBM had not complained about the drive. In fact, Mr. Irwin Rubin CMI's Chairman says that of the 90,000 drives shipped to IBM at the end of 1984, only 3% had been returned to CMI. Western Digital Inc. (Newport Beach, California) who supplies the controller for the IBM PC AT, reports similarly, that they had not received any returned controller or complaints from IBM about their product's quality either.

IBM's technical problem is one Mr. Dal Allen, president of ENDL Inc. (Saratoga, California) and author of the ENDL Letter, a monthly newsletter on disk drive standards and related topics, speculates is read-signal asymmetry. Asymmetrical read signals from a disk drive causes the VCO (voltage control oscillator) in the data separator to loose lock and the result is a read or seek error. A data separator synchronizes with stream of data bits coming from each sector in a track. It then extracts them from the surrounding overhead data that partitions sectors inside each track from one another.

The problem results from having the data separator physically located on the controller external to the drive. Thus, the data separator must be designed with extremely wide margins to accommodate a great diversity in operating characteristics from one disk drive to another. Controller manufacturers like Xebec Corp. (San Jose, California encountered similar problems in developing controller products for the IBM PC XT.

Having the data separator external to the drive was first implemented on the low-performance disk drive interfaces like that of the Shugart Corp. (Santa Clara, California) SA1000 and more prominently, the Seagate Technology (Scotts Valley, California) ST506/412. The rational for having the separator on the controller is that one controller can handle up to four drives, hence, to prevent duplicating the data separator on each drive, it is put on the controller.

What Mr. Allen and proponents of more intelligent interfaces like the ESDI (extended small disk interface) and SCSI (small computer system interface) propose is to adopt an interface that specifies the data separator on the drive. In this manner, the data separator is matched with the drive and the problem of read-signal symmetry is eliminated once and for all. However, Mr. Allen says that controller manufacturers are now solving the problem by designing data separators with even wider margins to accommodate the drive anomolies.

Having a poor 3rd quarter

Besides bad press reports about their disk drive, CMI also faced a disappointing third quarter at the end of 1984. As to the cause, CMI attributes at least $5 million of the loss to a one-time write down of inventory of older product lines and accounts receivable. Another part of the loss was blamed on misjudging the cost of materials and time for moving offshore, and the third on the unexpected low yields of the manufacturing line making the CM6000 family of drives--those shipping to IBM.

The company midjudged the cost of a critical component of the CM6000 drive, the encoder torquer motor. The company had originally estimated that the motor would cost $60 a piece. In reality they cost $100. On a drive that sells for $600 to IBM, $60 is nearly all the profit a company can expect to make. To correct the motor problem, the company has replaced the original motor with a lower-cost more readily available component.

Moving the printed circuit board manufacturing operation to Singapore took longer than anticipated, too. That production snag has since been overcome. Mr. Rubin is now predicting that the offshore facility will be shipping 30,000 drives by May 1 1985 and in July, 65% of its drive capacity will be produced in Singapore. In January, 1985, there were 500 employees offshore and Mr. Rubin expects this figure to double by year's end.

Banking on IBM

December 31, 1985 saw the end of the first IBM contract. IBM now had a very successful PC AT product that was already being stretched out in delivery over two months. They were pressing CMI to increase production significantly from the 90,000 delivered in 1984 up to 240,000 by the end of 1985, nearly a three-fold increase in production for one customer. However, the contract should produce nearly $150 million in revenue for CMI in 1985. But, with this large order, IBM now represents 77% of CMI's business.

In what Mr. Rubin calls delicate negotiations, CMI and IBM agreed to new terms for their 1985 contract. One detail of the contract was that CMI would receive a slightly higher price per drive. According to Electronics News 12/31/84, CMI would receive $550 per drive, however, the figure is closer to $600 according to a Data Storage Report source.

Another part of the letter of intent said IBM would provide CMI a loan. One reason CMI might have turned to IBM for the loan is that conventional capital sources have been dried up for a period. (Please see, Xidex-Dysan story elsewhere in this issue for more on this trend.) Selling more stock in a market where disk drive stocks are at a low point with the company's third quarter loss would not have been wise. According to the EN story cited above, the money, $6 milllion, would be used to improve manufacturing yields. Data Storage Report sources indicate that the figure is closer to $10 million.

Mr. Allen speculates on one way this yield improvement might be effected. He suggusts that CMI might well be matching controllers with drives. Thus, any drive that fails to work with an off-the-shelf controller is matched by hand with one that does. There was no confirmation of this practice from CMI. With the yield improvements, The company expects to be shipping 30,000 drives by April 1985. This production rate will produce the 240,000 units called for in the letter of intent and supply the company's remaining customers.

After IBM who's next?

One problem disk drive companies like International Memories, Inc. (IMI), (Cupertino, California), Seagate Technology Inc., and Miniscribe Corp. (Longmont, Colorado) is what business does a company pursue after it has had an IBM contract. Each were suppliers of the 10-Mbyte drive in the IBM PC XT. For IMI, the solution was to leave the disk drive business entirely. For Seagate and Miniscribe, the solution was to cut back on company operations until new customers could be found to replace lost IBM business. Both seem to have managed the aftermath well enough to continue operations.

Mr. Rubin believes that one reason IMI got out of the business was they had no follow-on, next generation product to attract the next wave of customers. The same is true of Seagate and Miniscribe. Neither had a 20-Mbyte drive product with an average access time of under 40 milliseconds. Both are currently working on such a product. However, the 20-Mbyte product is a mature product and most suppliers are already developing the next generation of products. There are two. One is a full-high product family with under 40-ms average access time. A second, in Mr. Rubin's view, is a half-high 20-Mbyte to 40-Mbyte drive family with under 40-ms average access time.

To shorten its development time, CMI purchased the IMI half-high product line for $1.5 million, including hard and soft tooling. Mr. Rubin said that CMI found the IMI mechanical design very attractive. Adding CMI's innovative electronics produced a high-performance high-capacity half-high next generation drive family, quickly. Mr. Rubin expects to have sample quantities of a 26-Mbyte (unformatted), half-high product with under 40-ms average access time in March or April of this year with production slated to begin in June.

One potential customer for this product is IBM, in their next generation AT product. However, an even better customer would be an IBM arch rival, like Apple Computer, Inc. (Cupertino, California), Hewlett-Packard Company (Palo Alto, California) or one that Mr. Rubin finds attractive, AT&T Technologies, Inc. (Lisle, Illinois). The one drawback to the IBM PC XT and AT is their size. Both take up a large amount of desk space. A competitive system with pc-compatible software, but in a much smaller package would sell well against IBM.

The ideal solution would be to use the smaller high-capacity, high-performance 3 1/2-in. drives. However, few with more than 20-Mbytes of capacity and under 80-ms average access time are available. (For more on this please see "Battle to supply IBM 3 1/2-in. Winchesters" elsewhere in this issue.) Moreover, such products will not be available for another two years. The only alternative is the half-high drives. Among the competitive half-high product, few offer under-40-ms average access time.

Advanced Storage Technology Inc. (San Jose, California) with its AST 96200 and 12330 Series offers a fast 20-ms average access time and capacities from 61.8 to 202 Mbytes, however, all require the ESDI or IPI (intelligent peripheral interface) interface. Moreover, the company has no track record in delivering large production quantities, a drawback when dealing with a large company like IBM or AT&T. CMI, on the other hand, does have a track record. It also has a drive that will operate with the ST506/412 interface.

Accessing CMI's current state

On the surface CMI seems in the best of health. The company has put behind itself a disappointing third quarter. It has signed a contract for all of 1985 with IBM, its Singapore facility is on-line and ready to ramp up production. However, these good signs belie the company's vulnerability. It still faces a lawsuit with Quantum Corp. (Milpitas, California) over the design of the drive CMI is selling to IBM.

It was filed August 2, 1983, the day Quantum was granted patent number 4,396,959. On April 2, 1984, filed a motion for summary judgment in Federal Court in Los Angeles, California. It seeks an early ruling on the original lawsuit. The patent covers a unique wedge-servo disk architecture which allows Quantum products a high degree of head-to-track positioning accuracy and data storage capacity at lower cost than drives using a voice coil actuator and closed loop servo system.

In addition, it faces strong competition from other drive makers hungry to bite into the IBM business. One of the two to challenge CMI is Seagate Technology (Scotts Valley, California). Its ST4026 product has specifications almost identical to the CMI drive being sold to IBM. A second threat comes from Tandon Corp. (Chatsworth, California). As described in separate story in this issue, Tandon announced it had a commitment from a major customer (the press have identified as IBM) for its TM703 hard disk family.

<BACK