18th April 1999 Archive
Browse by publication date, or search the site.
And why AMD is in a pickle
Just after last Christmas, adverts started appearing in the UK press for Celeron processors, with an egg-related theme. The ads may have run elsewhere than in the UK, and showed eggs in an eggbox, each branded with the Intel Inside logo. At the time, we dismissed this as just another bit of Intel tomfoolery but it has occurred to us since Easter that we were deeply mistaken. Chocolate company Cadburys now sells its Easter eggs all year round, although at one time sales were confined to a period just a few weeks before. The thin end of the Cadburys wedge started maybe 20 years ago, when a week or so after Christmas, its cream eggs started appearing in shops, much to everyone's shock. Intel sells eggs too, but as its Q1 figures show, they are golden eggs which hatch into gross profit margins of nearly 60 per cent. Just like chocolate eggs or real eggs, its chips have a limited shelf life and can go very smelly and pooey if they're left hanging around too long. Is this true or false? The answer is yes and no. For Intel, chips cannot be allowed to hang around too long. The geese in its fabrication plants have to keep laying new, bigger eggs. The end user must keep buying these eggs if margins are to stay high. That is not good for consumers, who do not necessarily need a 666MHz or a 1.2GHz Intel chip to run their copies of Encarta 2000. If you compare prices of the Pentium III now to the prices they will be in September, you can see how this Intel egg machine works. The graph here shows how Intel is practically halving the cost of Pentium III 500/400 chips over the next six months. Overegging the cake A couple of years ago, Craig Barrett, recently promoted to take over day-to-day duties from the poorly Andy Grove, outlined Intel's plans for the segmentation of the microprocessor market. The sub-$1,000 market, which he originally called the Basic PC market, would be served by the Celeron. This particular egg would be packaged in a Slot One configuration but without the cache associated with the Pentium II. At the time, many people pointed out that this was Intel sleight-of-hand. The Celeron is a Pentium II but with bits taken out, they said. But Intel did not want to be caught with too few eggs in its basket. It foresaw the growth of the sub-$1,000 segment a little too late and re-engineered its marketing and its packaging to sell the 14-carat version of its golden egg. To this end, it poured millions of marketing dollars into the Celeron platform, introduced products at cutthroat prices and continued selling the Pentium II at much higher prices. This pattern continues in the latest set of prices we have obtained from one of Intel's OEMs. Fee, fi, fo fum, I smell the blood of an AMD man When Advanced Micro Devices (AMD), filed its S-10 form with the US Securities and Equity Commission (SEC) a week before bad financial results came out, it was plain, as eggs is eggs, that Intel's pricing on Celerons and Pentium IIIs was hurting it badly. Unfortunately for AMD, it does not have as many geese laying golden eggs as Intel, and its prospects, geese factory wise, are also very limited. When Jack came back with the market with a few beans rather than money from the cash-cow, he got sent to bed without any supper and the useless beans got thrown into the garden only to sprout into a huge beanstalk next morning. Plucky little Jack eventually made off with the Giant's golden goose, and as he returned to the surface of the planet, undercut the beanstalk, pitching Intel to its doom. And everyone lived happy ever after. This market, however, is not a fairy story. The likelihood of Giant Intel falling off the beanstalk to its doom is slim. These are the market realities. AMD's K7 may be more golden than anything Intel has currently to offer, but Jack cannot produce enough of them to compete with the Giant. Worse, the chip giant has noticed AMD undercutting its share in the retail market and will continue to slash its Celeron prices and it will continue to erode the average selling price (ASP) of K6-IIIs and K6-2s. It will do the same with the K7 when limited quantities start appearing in July. Is AMD's very eggsistence threatened? When we were being roasted by the desert sun at the Intel Developer Forum last February, we were sitting behind one guy from Chipzilla who was chatting to a customer about AMD yields. If you're attempting to compete in the marketplace with a giant like Intel, the number of eggs you have in your basket is important. Intel has a huge basket and lots of eggs. They're not just desktop eggs but they're mobile eggs, server eggs and enterprise eggs. All of them make a profit and while some (like the original Celeron) are addled, most ain't. AMD hasn't got many eggs and its basket is tiny, though 300 or so people are wickerworking like mad. We couldn't help overhearing some of the Intel conversation about AMD yields and hastily scribbled down some figures. These, it must be stressed, are Intel projections and not Register speculations. K7 .18 micron technology will come on line in the first half of next year, produced at its Dresden Fab 30. This year, it will start producing .25 micron technology at its US Fab 25. Intel's thinking is that in the year 2000, it will manage over 5,000 wafer starts a week at Fab 25 and 1,500 at Fab 30. That is likely to move to over 5,000 at Fab 25 and over 5,000 at Fab 30 in 2001, meaning half of the wafer starts will be .18 micron. AMD may ship as many as 28 million K6-IIIs next year, but only four million K7s. The year after, the K6-III will virtually disappear with only eight million being sold, while the K7 will rise to 50 million. Full capacity at Fab 30, which will cost AMD nearly $2 billion when it is fully staffed up, is likely to be only 5,000 200mm wafer starts a week. Logically, it must therefore somehow find (borrow) the money for a new fab in the relatively near future. The Intel estimate, according to the overheard conversation, is that a .25 micron K7 will cost AMD $60 to make this year (including the cartridge), a .18 micron K7 will cost around $150 at the beginning of next year, but in the second half of the year, that will drop to around $60 again. These estimates, if they are true, mean that AMD needs a white knight to ride to its rescue. This is given that Intel has said it will continue to compete and compete highly aggressively in all sectors of the market. It is also a given that Intel will move early to .18 micron technology and its fabrication capacity is enormous, compared to AMD's. How much margin can AMD make on the limited quantities of K7s it can ship, while its shareholders bay at its door? While AMD has a deal with IBM Micro to fab up its chips, this also costs money. Perhaps, the only white knight that could conceivably save AMD's bacon is Compaq. This could be less likely now that Eckhard Pfeiffer has resigned as CEO. Another firm that has the means is Motorola. Eggs is eggs The Register does not take sides in the battle between Intel and AMD. We know from our internal sources at both companies that US executives view us as hostile. We are not football supporters and we do not consider that this is a battle between Jack the Giant Killer and the Giant itself, where we have to cheer on the underdog. Intel has 65,000 employees, vast financial liquidity, huge manufacturing capacities and excellent process technologies. Its revenues to the end of 1998 amounted to $26.273 billion, its net profit $6 billion, and its assets $31.47 billion. Its long term debt was $702 millions. AMD has 13,000 employees, limited manufacturing capacity and has been caught on the hop by its yields this year. AMD has a turnover of around $2 billion a year, and some apparently excellent technology in the shape of the K7. AMD's revenues to the end of 1998 amounted to $2.54 billion, its loss to $104 millions, its assets to $4.25 millions and its long term debts $1.37 billions. Quite obviously, AMD is no underdog. From the outside and from the outset, we have consistently reported instances where we feel Intel has taken unfair advantage of its size and position in the market. The US Federal Trade Commission (FTC) looks like it will endorse a settlement with Intel, which will not be good for the industry. But AMD has its own house to put in order, and the financial and fabrication facts speak for themselves. ®
Business 18 07:54
And yes, the pix are on a Japanese site
One swallow does not make a Spring here in the UK, nor does one picture of a daughterboard make a Cyrix 370 pin CPU. We said this would happen at the beginning of this year. (Stories: Cyrix decides to socket and see and Cyrix to go 370 pin in April) But over at Akiba, we do have a very nice and clear picture of a 370-pin socket on a daughterboard with wording on it which indicates support for both Intel and Cyrix chips. Other pictures at Akiba show that this daughter card, Intel PPGA370 Celeron to Slot One adaptors, show that it can use the Intel/Cyrix 370 pin chips on an existing Slot One motherboard. And a dual 370 configuration? Dunno about that guv. The Taiwanese upgrade daughterboard, marked 370 Socket Inside, will support processors from both firms. Intel and NatSemi, which now owns Cyrix, have cross licensing agreements on this sort of technology. We reported earlier this year that Cyrix would give 370-pin socket support. Other clone x.86 makers, including AMD, seem to be backing away from the idea, possibly because they're moving to Slot technology instead Try our search engine on the number 370 to see our numerous 370-socket stories. ®
Business 18 14:32
You can't publish Grateful Dead tapes on the Web, and you can't use the name - did Jerry die, or something?
You'll believe a trip can get even longer and stranger. Grateful Dead Productions and Ice Nine Publishing, custodians of the Dead's music and its publishers, respectively, seem intent on busting the crap out of the Deadabase Web site. Deadabase's sin is the publication of MP3 versions of live Grateful Dead concerts recorded by fans. As some of you youngsters probably aren't aware, the Dead famously allowed fans to tape and exchange anything they played. One would therefore expect - as indeed Deadabase reasonably expected - that these sessions could be freely shoved up on the Web and exchanged that way too. But no. According to an extremely fierce letter from Eric Doney of Donnahue, Gallagher, Woods & Wood, which Deadabase has republished on its site, "This firm represents Grateful Dead Productions ("GDP") and Ice Nine Publishing Co., Inc, publishers of the music of the Grateful Dead, in connection with investigating instances of illegal duplication and/or distribution of recorded music. We have reviewed your Deadabase.com site in which you offer for download at least forty (40) live shows apparently recorded by fans of the Band [sic. Lawyers, dontchalovethem? What would fans of The Band be doing at Dead gigs?] over the years. As you are no doubt aware, for many years, the Grateful Dead have permitted their fans to record performances for their personal enjoyment. Individuals who record the tapes are also permitted to trade them among fellow Deadheads. No one other than GDP, Ice Nine and their expressly authorized agents, however, is permitted to offer the download of recordings of Grateful Dead performances under any circumstances." (Our italics) The letter doesn't go on to say under what conditions GDP et al will offer the download of recordings, but one could surmise that folding green stuff might be connected in some way. We of course would be the last to suggest that their streets run deep with poisoned wine or their doorways crawled with fear. But for Deadabase, it gets worse. The use of the name Deadabase is claimed as a trademark infringement, and the site itself "contains a great deal of Grateful Dead material and could reasonably be construed by a user to be sanctioned by my client... We ask that you eliminate any recordings of the Grateful Dead from your site and that the use of any derivative of the name "Grateful Dead" also be eliminated from the site." Depending on whether or not Deadabase is making money out of the recordings (it says not), then the MP3 part of the action may have justification -this. However, is gratuitous. But they say they've got a warrant - do we figure they're gonna come in? Deadabase thinks (hopes) that the move hasn't been sanctioned by band members, and claims to have spoken to song writer John Barlow. It claims Barlow has been shown sales figures that show Deadbase isn't making money out of the recordings, and that Barlow has indicated that the band wants fans to be able to exchange MP3 versions of its recordings on the Web. But pending developments, Deadabase has pulled the recordings. ®
Business 18 16:19
Let the hyping commence - but how finished is 'finished?'
After months of talking down expectations for Windows 2000, Microsoft started talking them up again on Friday, confirming the suspicions we expressed just a little earlier (MS starts Win2k push) that this time Redmond was definitely going for a Win2k launch big-time. Senior VP Jim Allchin, the man responsible for the rollout, has been schmoozing PC Week Online, and promising that Win2k beta 3 will be with 500,000 beta testers next month. Actually, our spies tell us that beta 3 has been available on the MS Developer Network for over a week now, and that at least four hotfixes have already been posted there. Don't even try to look though - it's padlocked. Allchin's appearance on the interview trail is the latest sign that the rollout has started already, the other two obvious ones being the engagement of the Corporate Partnership Program (CPP), designed to get corporate users started on an upgrade programme that leads from beta 3 to gold version, and the OEM programme, whereby OEMs will ship beta 3 machines plus an upgrade certificate. What was missing after Microsoft moved on these two was the widespread beta, and Allchin's promise of 500,000 copies seems to deal with that one. A lot of people are going to get their hands on Win2k, and they're going to get it Real Soon Now. From what our good friends at ZD report, Allchin seems remarkably perky and optimistic about the project, not coming up with any actual ship dates for the gold product, of course, but it's worth noting that our good friends now refer to Microsoft internally expecting shipment in October. This is a little bit different from previous suggestions that selected customers might receive some form of pre-release version of the product in October, so we should watch out for bits missing from anything that ships earlier than late November. We should also watch who receives anything that ships in October. Allchin says Microsoft will continue to update code until Win2k ships, and depending on what he means by that it could have a severe impact on the product actually getting to the PC companies. They need Microsoft to freeze the code, send it to them and then give them time to test it before they ship, so the further Allchin rolls with his update programme, the later the PC outfits will be. Microsoft also needs time to cut CDs. Will "ship" not really mean ship? You read it here first. Allchin also reveals something rather interesting about the 'security' of Win2k. The OS now includes protection that stops applications from replacing core system DLLs. This obviously is good if it stops rogue applications trashing your system with their install procedures, and it's certainly about time Microsoft did something about the DLL hell its faulty admin procedures create, but it's potentially a control system too. If an application requires the replacement of a system DLL, and if only a Microsoft patch or service pack is allowed to do this (one presumes), then Microsoft gets to call the shots, doesn't it? One of the pieces of software that may be threatened by this procedure seems to be, spookily, Novell's NDS for NT. ®
Business 18 17:46
Add those to the i840 spex we leaked...
The good ship Intel is developing leaks across the world and the latest holes in its hull are the 810, 810e and 820 chips specifications. We have already revealed i810 details in an upcoming processor here, and have also published information about the i840 workstation chipset and two motherboards here. Go to Kbench where you'll find some very juicy info about the 810, 810E and 820 chipsets. The English language tables show how these chipsets relate to future PC platforms, and the information appears to be genuine. AGP Pro info has been around for a while and if you go to Kbench, you'll see all sorts of details about the AGP Pro slot specification. The information is in Korean but the diagrams are in English. There is an English language version of the AGP Pro specs only here. Check out The Register tomorrow for more information we can offer you from our own sources about the 810, the 810e and the 820. ® See also Celeron Slot One dead Celeron to go 100MHz FSB
Business 18 22:19
And CFO Earl Mason heads for the hills
Compaq CEO Eckhard Pfeiffer has fallen on his sword. He resigned today, along with CFO Earl Mason. Chairman Ben Rosen will look after the shop along with vice chairmen Frank Doyle and Robert Ted Enloe - until a new CEO comes on board. Pfeiffer's departure looks like palace coup: remember, he got the job in the first place by persuading Rosen to dismiss Compaq founder Rod Canion in 1991. Pfeiffer clearly lost the confidence of Rosen, the man who supplied the initial venture capital for Compaq and who -- it is now clear -- remains the most important man in the company. With Rosen's backing, Pfeiffer could have toughed out the recent bad press and share performance, following a spectacularly inept handling of the company's recent bad quarter. But -- and its a big but -- Rosen has been dissatisfied with Compaq's financial performance for two years, according to the Financial Times. Rosen thinks the company's strategy remains sound -- unlike in 1991 when he sacked Canion. But execution is poor. This is somewhat ironic, considering Pfeiffer was never known for the Vision Thing. Pfeiffer was promoted to the top to re-engineer Compaq as a leaner, more efficient company. Under Canion, the company had grown fat and flabby - and expected customers to pay up to three times more for Compaq PCs than from equivalent models from new competitors. Compaq is fat and flabby again -- despite Pfeiffer's much vaunted obsession with supply chain management. Too much management energy has been spent on the Digital takeover -- and the company is paying the price for its introspection. Compaq is also paying the price for its dependence on the channel. In the early eighties, this was the secret of its success. In the Internet-savvy late nineties, the channel is becoming a millstone round its neck. We expect the new CEO -- whoever he or she is -- will accelerate the transition to a direct-mainly product sales strategy. The new boss/Rosen lackey will have a tough brief to follow: "I hope never again to issue a profit warning," Rosen says. Rosen will unveil the company's Q1 results in all their gory detail on Wednesday. For last week's profit warning, Mason is also walking the plank. Although the timing could just be co-incidental. Mason is off to be a CEO in another company outside the sector, Compaq says, so he has had his escape route planned for some time. ®
Business 18 22:36
- Comment Renewable energy 'simply WON'T WORK': Top Google engineers
- Useless 'computer engineer' Barbie FIRED in three-way fsck row
- Game Theory Dragon Age Inquisition: Our chief weapons are...
- 'How a censorious and moralistic blogger ruined my evening'
- Leaked screenshots show next Windows kernel to be a perfect 10