16th > October > 1999 Archive

The Register breaking news

Intel'sh Cashcades to cash in on cache inshide

As we reported last Sunday, Intel will introduce its Xeon Coppermine processors with 256K of cache on die on the 25th of October next. These were formerly codenamed Cascades, or Cashcades as we prefer it and will neatly sit in Slot 2 of 840 mobos, as revealed earlier. We will see 512K and 1Mb versions of these chips with much higher clock frequencies probably in Q1 of next year, as Intel tweaks its .18 micron (Coppermine/Goldmine) technology. The Celeron III will also benefit from this ramp and have 128K of cache within the processor. The server and workstation Xeons are the so-called Cashcades (shurely shome mishtake, ed), promised for many a year and are intended to dominate both the server market and the workstation market. More significantly, the i840 is also aimed at the high end gaming market. This is Intel's short-term challenge to the Athlon phenomenon. A number of large PC manufacturers and third party motherboard makers will roll out solutions based on Cashcades techology, including, as we reported earlier ths week, Intel itself with its Outrigger (OR840) mobo. Sources close but friendly to Intel tell The Register that the i840 chipset was not designed by the same gang that is currently walking the tightrope for the i820 debacle. Unlike the i820, the i840 Carmel will use dual channel Rambus and is a different kettle of fish altogether. Raghu Murthi, marketing director of Intel's workstation division at its Dupont plant, would not, when we met him at a briefing yesterday, publicly comment on the i840 prior to its launch. He did confirm, however, that it would happen in Week 44, which by a strange coincidence is exactly the date we have for Cashcade Xeons. Meanwhile, sources intimately close to an Intel engineer have told The Register that the company is happy with Coppermine benchmarking internally. The firm has made many architectural changes to the original design other than just to the level two cache, and disclosed some of those changes at the Microprocessor Forum two weeks back. Our reliable source, who would lose her job and her boyfriend if we quoted her directly, said that Intel's Coppermine processor is beating Athlon K7s even when the latter are clocked high. Any initial glitches, she said, are now ironed out and frequencies (MHz speeds) will now ramp very much faster than Intel expected itself. We're still unclear about exact details and pricing of the Coppermine Celeron III but reports from yet another set of reliable moles suggest that Intel may use its 7 November Celeron price change date to introduce a Coppermine 128K version. If that isn't the date, then it's likely to be not long after. Again, Intel will rapidly ramp these Celeron IIIs at speed. Meanwhile, one of these moles (can't remember which one now) points out that Kyle Bennett's benchmarks over at HardOCP, show a SiSoft Sandra screen describing a tested Coppermine as a Celeron III. This, we are given to understand, is not completely a coincidence. Tomorrow, we will take an indepth look at Intel's roadmaps for the Year 2000, according to our best leaked sources. ® * Factoid. A wag suggests that if Intel keeps knocking RIMM slots off the 820 mobos, it will eventually produce a Vancouver mobo with no Rambus support at all, thus solving all of its problems. Another points out that Necx is selling Intel Vancouver and Cape Cod mobos, but without being able to give a delivery date. Do they have two RIMM slots, he wonders? See also Xeons to beat desktop PIIIs in MHz stakes Satan's little helpers get hammers out to i840 mobos i840 Carmel mobo details leak Celeron III, anyone? Athlon 700 prices to crash: AMD, Intel war escalates Coppermine: it's an Intel goldmine
The Register breaking news

Apple reverses direction on G4 downgrades again

Apple seems to have a real problem making up it's mind, these days. First it announces that anyone who has already ordered a Power Mac G4 but has yet to receive it will now have to take a machine with a slower CPU for the same price by re-placing their order. Apple's move was announced along with the company's Q4 1999 results. The company said that because of problems with the supply of 500MHz PowerPC 7400 (aka G4) CPUs (a tacit acknowledgement of the bug that prevents the G4 from running at 500MHz or more), it was downgrading its entire G4 line by 50MHz -- so what started out with 400, 450 and 500MHz machines now comprises 350, 400 and 450MHz models. In response to the massive outcry from understandably disgruntled customers, Apple admitted its gaffe and its direct sales force yesterday began emailing buyers that orders for 400MHz and 450MHz machines will be honoured, and that orders for the 500MHz version will be converted either to a 450MHz machine for $350 less, or a 450MHz for the same price but with an extra 128MB of memory. Now, less than 24 hours later, the Mac maker has changed its mind again and stated that all deals are now off, and buyers will have to re-order new, slower machines after all. An Apple spokesman told Web site MacWeek: "We are honouring [advance] orders from just a small number of individuals." He said that the email from the AppleStore had been a mistake since the production process made it impossible to fulfil all but a few now non-standard configurations. So much for Apple's build-to-order system, the whole point of which is to produce non-standard configurations. If it can't delay orders for old-configuration models and re-route them through the BTO channel, it's a very poor do. Meanwhile, US dealers contacted by MacWeek said they too were now being forced to contact customers to tell them their existing orders can no longer be fulfilled. Said the Apple spokesman: "We are apologising for the inconvenience this has caused our valuable customers. Fortunately, most of our customers understand the need to do this to meet demand, and almost all of the customers we have heard from are reordering." For Apple's sake, we hope so, and that it can quickly put this colossal balls-up behind it. ®