Two years ago: Register guddles trout while Intel goes fishin’
And talking about the First Amendment...
Intel has some interesting things afoot but surely one of the most intriguing is what to tell its end users and its OEMs about the Klamath and Deschute designs up-a-coming. Two British magazines were brave enough to stand up to threats emanating from Intel's legal department last month and published information about its plans in the face of a possible intellectual property suit. (This would have been a test-case in the UK and Europe - possibly around the world, can journalists be bound by non-existent NDAs - anyone remember the Spycatcher case? Secondly, the info was also published on the Web). One of the problems Intel undoubtedly has, is that Taiwanese motherboard manufacturers are a little leery about licensing the new slot technology. After all, will they put themselves in the hands of a monster company with perhaps, might we suggest, monopolistic intentions? Cache companies have also expressed concern to The Register that the new designs could place them in a similar position. But possibly the more intriguing question is how Intel does the magic trick of changing pin outs on Pentium Pros and still keeping its "loyal" user base. OEMs at several large companies have confirmed the truth of the update Intel gave them in October. The favoured few (hi Dell, hi Gateway) are likely to have Klamath introductions in mid year. There will be two offerings, one a 266MHz/512K card with error correction and the other a 233MHz/256K without. Intel is also offering two thermal solutions for the Klamath, one with an ATX style heat sink and the other using a so-called LPX/L solution. The chipset at introduction will be 440LX based. How, then, does Intel persuade end users and OEMs they should punt on the Klamath river? Intel will attempt to push the argument that the Pentium Pro 200/512 offers the best performance for quad systems while the Klamath is suited to the low-end server market. The other way to persuade infidels is that the Pro will have greater volume than the Klamath in the first half of 1997, although that will change in the second half of the year, with the Pro ramping down. The Q1 card price of the Klamath will be $725 for the 233/512 offering, a price that includes the Pro chip, tag memory and 117/133MHzburst synchronous memory. The Klamath will be a commodity offering. Intel is very likely to keep pricing of its Pro 200, 180 and 166 chips stable when it releases its next prices in February next year. That will give some stability to the Pro market as it introduces the Klamath, which will, of course, have MMX extensions. The Klamath design will pave the way for Deschutes, which comes in Slot 2 configurations at the beginning of 1998. Intel dubs this a "faster Siskiyou" meaning it will use a similar chip design with a few relatively insignificant changes for bus ratios between 266and 300MHz. It will also have a 64-bit GTL and memory bus. While the Klamath is intended for 1-2 processor designs, the Deschutes is aimed at 1-4.Clock speeds will be a maximum of 266MHz for the Klamath and300MHz for Deschutes. Both will come with 16K level one cache, both will use the MMX extensions but while the Klamath will initially be aimed at the volume desktop, the Deschutes design will be targeted at the high end. The differences in chip set design are that the Klamath will use the 440LX chipset while Deschutes will use 440BX chipsets, whatever they are. Eventually, the Slot 2 Deschutes design will offer as much as one to two megs of cache memory and will be based on 450NXtechnology. Other OEMs The Register >has discussed this matter with tell us that Intel will eventually introduce notebook technology based on these daughter cards designs. One thing the pesky OEMs and other people in the industry would like to know from Intel but have fat chance of getting any answer to is why the company still turns in margins in excess of 65 per cent while everyone else in the industry has seen their margins shot to shreds over the last two years. We think we should be told... ® From The Register Issue 38, 15 December 1996
Sponsored: Hyper-scale data management