Several years ago, a charming, enthusiastic but somewhat monomaniacal lawyer tried to talk us into buying a building in New York City. But somewhere between Location, location and location and Mother Nature only made so much land we grew edgy. We weren’t particularly worried about the wisdom of owning our own little piece of the […]
Several years ago, a charming, enthusiastic but somewhat monomaniacal lawyer tried to talk us into buying a building in New York City. But somewhere between Location, location and location and Mother Nature only made so much land we grew edgy. We weren’t particularly worried about the wisdom of owning our own little piece of the Big Apple, but we just didn’t feel comfortable paying the market price for the chunks of New York we were offered. In the end, we stayed put and at first suffered greatly. Real estate magnates kept getting in the headlines as their net worth nearly forced newspapers to switch to scientific notation. At this time – the middle of the last decade – any idiot could amass a paper fortune in property and many did. Only a couple of years and a stock market crash or two later, the real estate moguls turned out to be somewhat overcommitted. Now the mightiest property empire in the world, Olympia & York, wobbles between partial and total collapse. O&Y has just begun to learn what others long since found out: real estate is a game of crowds. When the mobs rush in, no rent is too high. When they desert, no rent is low enough. Even the most violent fluctuations in real estate wouldn’t matter much if property owners had patience and used their own money. In the long run, all but the most foolish investments would pay off. But real estate, for most of the high rollers, is a leveraged undertaking.
During the market’s rise, players used the increasing theoretical value of their holdings to double and redouble their bets. All this would have been fine except for the one step overlooked by the paper zillionaires: selling out before the peak. That means that you have to recognise that special moment when the last, big sucker shows up ready to buy for maybe a small discount every hyperinflated square foot. The signals of an impending peak may sometimes be hard to detect, but in the case of O&Y there was plenty of warning. If you don’t believe us, ask Donald Trump. But don’t ask IBM. IBM is suffering from a real estate collapse of its own, but hasn’t quite come around to understanding it. The real estate losing value in IBM’s empire isn’t built on soil but on sand – silicon and glass. IBM’s troubled real estate empire is made of circuit boards, not two-by-fours. But the forces that have given it a grossly inflated value all too closely resemble those that led investors to fight for an interest in Canary Wharf and the other extraordinary properties bought or built by O&Y. Recently, mainframe shops have taken another look at IBM-compatible memory… and found that it works both technically and financially. The immediate impetus to this movement among 3090 users has been the widely publicised effort by IBM to stamp out traffic in remanufactured memory boards. As IBM swooped down on parties whom it accused of passing off rebuilt boards as authentic IBM ones, some popular blocks of genuine IBM memory 128Mb units, for instance – became temporarily scarce and expensive. Hard-pressed customers who needed memory found that they could no longer afford the IBM brand and were forced to consider IBM-compatible alternatives. These customers have discovered that compatible memory, like IBM’s own, is very reliable.
By Hesh Wiener
They still don’t know how the residual value of IBM-compatible memory will compare to that of the real thing, but they suspect that the non-IBM product will depreciate more rapidly. They may not be correct; it depends on the kind of IBM memory you are talking about. During the 3090 generation, IBM manufactured four processor families: base, E, S and J. It also made three generations of main memory. The first and slowest technology is called Night Train; it was used in base machines. The same components are also found in the expanded memory of E systems. Then came Eagle, a faster type of memory used for central and expanded storage on E machines and expanded storage on S processors. Finally, IBM produced Antelope, used in J and J-prime machines. Any IBM 3090 can use faster memory than I
BM originally supplied, although it will not improve system performance. Our contacts in the leasing business believe that all the unauthorised remanufactured memory used Antelope components. So even if, as IBM says, the resultant boards are subject to failure, they are inherently faster than many of the memory boards IBM sold… and they will work (technically, not legally) with any 3090 or J-prime processor. Customers buying memory upgrades from IBM may not always get Antelope. For example, a customer upgrading main memory on a 3090-E might get Eagle technology; that same customer adding expanded memory could get either Night Train or Eagle. IBM’s price for memory is consistent regardless of technology… but customers reselling 3090s will find that the used memory market distinguishes among the various generations. Consequently, customers buying central or expanded 3090 or J-prime memory from IBM (or anyone) should try to get Antelope technology rather than slower circuits. While Antelope memory, particularly large-capacity increments, will almost certainly have higher residual value than IBM-compatible boards, Night Train or Eagle memory may turn out to be worth less than faster boards made by IBM’s competitors. And even the Antelope technology could lose market value much faster than previous generations of IBM main memory. IBM mainframe memory is very costly compared with memory for other machines – $3,200 per megabyte versus $371 to $844 on various IBM AS/400s or $300 on an RS/6000. Granted, much of the circuitry and packaging used by IBM is very sophisticated. But the underlying memory chip technology is pretty much the same across the industry… and it is priced, at wholesale, well below $40 per megabyte, which allows IBM and its competitors (not only memory makers but also systems vendors) to undercut IBM and still make a profit. Thus, the real estate on IBM’s memory boards is very costly indeed. It is not as easy to compare IBM’s disks with alternatives, but that will surely occur as array technology becomes widespread. The standardised processor benchmarks sometimes used to match IBM’s mainframes against other kinds of computers aren’t really worth discussing. Nonetheless, customers know that memory isn’t the only part of an IBM mainframe system that is priced on the high side.
For the birds
They realise that they are paying a premium for their IBM mainframe systems… and they are not as comfortable with this as they once seemed to be. It is still chic to house corporate data in an IBM mainframe, just as it is still nice to have your company in a prestige building. However, IBM’s future depends not on customers’ wishes but instead on their behaviour. Unless mainframe costs fall sharply, we think sentiment will turn. As in the property market, the problem won’t be limited to some unrented office space. If a lot of companies move applications out of mainframes at once, IBM – and every equipment owner will find that its situation, like Canary Wharf, is strictly for the birds.Copyright (C) 1992 Technology News of America Co. All rights reserved.