You missed the real cause of my jubilation - that I got the capitalization right almost 2 1/4 years ago.
The IT Jungle story still doesn't manage it. But an invitation IBM recently sent to analysts says "... the eClipz processor ..." YES!!! You wouldn't believe how much time was spent on that. You sometimes have to wring clues out of the thinnest information, and the capitalization of a mnemonic can be a major clue to the importance of its components. The IT Jungle piece seems to be a combination of four sources, three of which were reasonable. One was mine, and one was one derived from mine. It's a bit like analysing a commentary on the synoptic Gospels. He's not quite right on many details, but the crucial one is the discrepancy between the z9 ---> z6 native cycle time improvement and the delivered grunt of around +50%. That means this machine works differently. Not worse, not better - differently. And I don't think he understands decimal arithmetic. "Money math". The greatest strength of System/360 even at launch - how does he think it got where it got? ZAP was and is a wonderful instruction, if you were used to what went before. The piece is overcrowded with numbers. Speeds and feeds. What matters is what comes out the back - I've long since stopped caring how it's done. I lack the qualifications to judge design decisions like cache sizes - if you see the guys that make these decisions working, you leave the room with your head spinning. Serious, serious math. Sometimes it gets funny. 1,199 signal pins and a total of 8,765 pins. So what do the other 7,566 pins do? Knit? As I've said here before, I believe it would be a good idea to prepare for some turbulence - similar to but greater in magnitude than the issues we got with the 128 byte cache lines. I'm really not that happy with the implied reduction in SMP factors. I've heard the opposite in some rumours. As I've also said, I do not doubt for one second that IBM will meet the overall box performance target. But I think it would be very prudent to ensure that you can support performance measurements at fine granularity - transaction level, subroutine level - very rapidly if asked to do so. Who markets Strobe these days? Stick a few bucks in the stock. Again, it's the old law that the grumblings of one unhappy user can drown the cheers of 99 happy ones - except this time I'd expect two unhappy users. I don't buy the "z9 sales affected by z6 leaks" angle. In the first place, there have been no substantive leaks. And in the second, the z9 is very much a known quantity and the safe bet. I would not be at all surprised to see z9 sales pick up slightly in 2008Q1. Anyway, next week Charles Webb is going to read his PDF to those analysts too stupid to have found it for themselves. Which is quite a few. I won't be taking part - the bit I miss most is the Q&A at the end where each analyst spends the first 75% of their allotted question time gushing to the executives. "I mean. wow, I'd just like to say how wonderful this is for our customers ..." Get OUTAH HERE, you moron! Anyone with a brain has known this stuff for two years! You frankly wouldn't believe it. "Hi, I'm so-and-so from household-name." And then the dumbest question you've ever heard. I'm sometimes amazed that you can't hear the executives smirking when they answer. On at least one occasion a few years back a question was asked directly of an executive - there was a slightly muffled noise and the facilitator came back with: "Well, I'll ask xyz to answer that one instead." I strongly suspect the original target was rolling on the floor with several colleagues sitting on him, trying to stifle his paroxisms of laughter. -- Phil Payne http://www.isham-research.co.uk +44 7833 654 800 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
