Guy Kewney's Diary

This time, it's sheer bloody-mindedness by inanimate objects. To be specific, it's the office switchboard. It's busy telling anybody who calls that I'm 'engaged', BUSY, occupied. I have no way of knowing this, and by the end of the day, when the penny finally drops and I call myself on my office phone by dialling the number on my mobile, I have destroyed several valuable friendships by open disbelief when people claim to have been trying to call me back. "I can't be engaged! I've got bloody voice mail, you fool!" Only less diplomatic...

The problem, very simply, is that "Trends" for PC Mag has to be written by Monday last week. A spreadsheet designed to produce the deadlines has obeyed the old, trusted GIGO (garbage in, garbage out) law and on being given the wrong dates as data to work with, has produced a schedule which will produce the March edition in March. That would NEVER do. It's due out on Feb 3, for goodness sake.

What actually seems to have gone wrong, is a conspiracy by hardware and organisation. The switchboard, which was pllaying silly games last week, was taken down for the weekend for fixing.

As a result, my dial-in access to the network didn't work. As a result, when I got in, there was a weepy message from The Vampire, pointing out that he had a busy week and had hoped to get all the copy edited over the weekend, and he'd come into the office to edit copy. And there was no copy to edit. What I don't understand, is how I might have been expected to get into the office over the weekend (assuming I'd been silly enough to try). I mean, some of us can transform into bat form and come in via the air conditioning duct, sure; but not me.

Tuesday

Sun. It has the unluckiest day ever. It's launched a really neat series of desktop machines. All Unix-based, all PCI bus, and priced at a level which challenges Dell and Compaq in the "workstation" market. And it should make it into the Trends story, but it didn't, because it turns out that the schedule has been moved forwards further; the page is closed. Bah.

The real problem, of course, is that Sun's UltraSparc processor is a totally closed book to most PC pundits. They've never used one, don't know what it's like. And that includes me. And believe me, it's not for want of trying. At today's launch at the Institute of Contemporary Arts building off Pall Mall, we play a strange, ritual game. It's antiphonal, and versicle and response are quite formalised, as follows...

Sun High Priest: "These are truly wonderful examples of technology."

Journalist Worshipper: "We'd really like to see them."

SHP: "You should get our marketing people to give you one on a long-term loan."

JW: "That would indeed be a good idea. For most PC people, it takes more than just a couple of days to get to grips with Solaris. We tend not to know our way around Unix."

SHP: "Oh, well; we'll certainly be able to help."

And then the two part, and both dismiss it from their minds. Sun simply doesn't yet understand the desktop market.

Mind you, these new machines might lead to a learning experience inside Camberley. Priced at around £3,400 for a Pentium II-level machine fully fitted with workstation level disk and graphics and software, it compares pretty closely to Compaq's offerings. If things stay as they are, then Sun will certainly have to double production, maybe more, to cope with demand from people who would otherwise desert Sparc for the PC standard.

But of course, things aren't going to stay as they are. Hewlett-Packard is cutting its prices, 17 per cent, or so, and we're only a couple of weeks from the next Intel chip announcement -- probably, Deschutes, which is a Pentium Pro level version of the Pentium II -- usable in multi-processor boxes. But Intel will be talking about Katmai, soon. Watch this space.

Wednesday

For portability, you have to sacrifice something. When I selected this ThinkPad notebook, I went for a model without a CD. And now Quake II has arrived, and I want to install it. And there's no CD.

A friend (ha!) tells me I can use the CD on my desktop. I point out that the ThinkPad has no network card. "Not to worry, Guy! Install Direct Cable connection -- it's in the start menu under Accessories."

It begins. First, I install it on the desktop. It says it needs rebooting. I reboot. Then I plug the two machines together with a Laplink parallel cable (it is supposed to work). Then I run the Direct Cable Wizard on the notebook. It says it needs rebooting, too. I reboot. The two machines ignore each other.

We start the troubleshooting wizard. It asks me subtle things like: "Have you set one machine to 'host' and the other to 'guest' or are they both host or both guest?" -- a dumb question worth asking, I suppose, but right now, calculated to infuriate me. After working my way through the entire list, it tells me I've reached a problem which the TroubleShooter can't help me with. Acall to my friend, then.

"You did set up Microsoft client for networks on the notebook?" No, actually, I hadn't. Silly me. Install client. The two machines ignore each other, despite two reboots each. The host now asks me what domain I think the guest ought to be. Domain? this is two Windows 95 machines! Another call. Nothing.

And then the notebook screen goes dark.

The day ends in fury. This wretched ThinkPad is small and cute and fast and has a huge disk; and it's dead.

The reason it's dead couldn't be simpler; the battery is flat. And the battery is flat because when I came home last night with the thing in my bag, I brought the 'power brick' with me, but -- naturally -- I didn't bring the cable. Why carry a perfectly standard mains cable around?

Because IBM designers are a bunch of addle-headed, half-witted, idiots, that's why. In a world full of standard tea-kettle cables, it was bad enough when Toshiba introduced the two-wire ghetto-blaster power cord. But at least I can go into any music or record shop from here to LA, from Hong Kong to Auckland, and buy a new ghetto-blaster cable. What I can't get, is an IBM ThinkPad cable, 1998 edition. IBM has invented its own standard. It's like a ghetto-blaster calbe, but it has three (count the blighters!) cables.

If you want three cables, the tea-kettle lead serves. It has served faithfully for IBM users for years. True, it's a high power design, and most portable PCs are low power devices, so it's over kill. But if you want small, you use the ghetto-blaster, right? No: you invent a new standard.

You cannot find an IBM ThinkPad power cable in any shop anywhere, and in something of a temper, I rang my contact at IBM. "I dunno," was the summary of his response. "It might have an explanation, but I've never found anybody prepared to reveal it," he admitted. "We get more complaints about this than anything else."

Thursday

Want to know when the 64-bit Intel "Merced" chip will ship? Got any idea what "Katmai" is? A massive clean-up of the hard drive; this (dated September) pops out. It never got printed, for some reason: it's the "secret" projections of Intel's chip plans for the next two years.

The first authoritative projection of Intel's plans for its 64-bit Merced chip, made by Microprocessor Report, show that Intel is expecting to be able to move the industry beyond the X86 instruction set before the millennium.

The Merced joint development with Hewlett-Packard, says the report's 1997 "Microprocessor Forecast", is one of two seventh generation processor products. MPR analyst Linley Gwenapp expects the first Merced shipments no earlier than 1999. And hot on its heels, in the same year will come "Willamette" -- an 800 MHz "dinosaur" of a chip, far faster than Merced at X86 code, but about half the speed of Merced if Merced is allowed to run in its new native mode.

The forecast, widely acknowledged to be the definitive industry reference on Intel's plans, reveals several processors which Intel itself has never publicly mentioned. In particular, the secret Katmai chip, due out next year or early 1999, will introduce MMX2, aimed -- perhaps somewhat late -- at enhancing 3D graphics performance.

For PC Wintel-software followers -- the PC industry as a whole, that is -- the good news is that with Willamette, Intel believes it has cracked the problem of producing a fast x86-- even while it is working as hard as possible on replacing it.

One of the disappointments of the RISC processor dream has been the discovery that it did not provide a magic gateway to leapfrog the Intel X86 architecture. Motorola and IBM and Apple, in particular, are in a disarray caused at least in part by the utter failure of the PowerPC architecture to eclipse the Pentium II.

But matching that disappointment, has been the slowdown in x86 improvements. Speed of Pentium II family processors has grown no faster than clock speed -- and sometimes, less than that.

Willamette, however, breaks that pattern. Gwenapp's Forecast confidently reckons that the seventh generation processor will run 30-50 per cent faster, at the same clock speed, than the chip which will be the industry standard Intel part in 1999. That, Gwenapp believes, will be Katmai -- an enhanced version of the Deschutes chip.

Deschutes, which is expected to render the Pentium Pro obsolete, brings the 100MHz system bus to the desktop, and the Klamath architecture to multi-processor (four-way) servers. It also takes the Pentium II or P6 family down to the notebook arena for the first time.

Katmai will take P6 processing up to its clock limit, believed to be around 500 MHz.

Key to Gwenapp's projections, is his indepth analysis of Intel's silicon foundry capability. To put it crudely, he knows what it costs Intel to produce a wafer, and can work out what Intel will have to charge for the chips on it, as designs get more complex, and as silicon features shrink. What he does not attempt to do, is analyse whether the world will want to buy what Intel designs. So, if (for example) there is going to be a collapse in the PC market in favour of simpler Network Computers, his projections won't show this.

On the other hand, Intel has dominated the PC industry for over 15 years, and in every month of those years, survival has depended on second-guessing its plans.

Key to its plans, are its silicon capacities, which Intel does not publish. Gwenapp's Forecast gathers together the most comprehensive imaginable analysis of what is known about Intel's capacity, its plans, and its potential ability to carry them out. For further details, see www.mdronline.com.

Friday

Direct Cable connection it turns out, is a product designed for show, not use. Not only has nobody at IBM been able to sort me out with a hint, but neither is Microsoft able to assist. Which wouldn't be the end of the world, but for one thing: in the process of installing DCC, Windows has taken the opportunity to throw away all my Internet connectivity data.

So if I run the browser, it first tries to dial a modem port at MSN, then an alternative port at LineOne. And then it says "Dialup networking is incorrectly installed." Funny; it wasn't yesterday. But I'm an obedient servent of the machine; I do as instructed. No use; so in the end, I remove the dialup network adapter from the driver list. And so back to the traditional Ziff-Davis connection, through our megabit pipe onto the LAN? Wrong. Windows Networking has gone into a fit of sulks, and has thrown the teddy bear out of the pram. All my DNS numbers -- gone. All my ISP numbers -- gone.

Normally, I'd run for help to the Online Tossers. They are not amenable to requests for assistance, because &deity& has struck, introducing a power cut into the office right at the moment when they were bringing up the backup Internet server after a crash on the main one.

It's a stupid way of spending a day, when I ought to be chasing Traveling Software for news on the "black screen" problem. Some display card makers have decided that Windows-standard hardware is for wimps, and have started ignoring the Windows GDI calls. The result is that when I dial into the office (assuming the wretched phone is working) the laplink connection works, but the remote control window shows all the GDI calls -- which are not there. So it's black.

My card is an STB Systems Nitro 3D, and if it were as fast as STB thinks it is, it would still be useless. To run it (as I mentioned last year) I have to set the Gateway into SVGA mode -- 16 colours, 800 by 600 pixels. And the fact is, SVGA mode is something NOBODY uses; and it is full of nasty bugs. Notes, for example, fixed the "sticky keys" bug years ago -- but not in SVGA mode, and so my cursor keys, instead of moving the cursor around the screen, move the screen around.

I think I get the hint. Time for a weekend. Maybe I can do something pleasant and relaxing, like digging up the back drain, which is blocked and leaking a gallon a minute into the cellar? And a nice Saturday to you, too.