SUMMARY Pentium II family dead, Pentium III/450 sent to gulag, fresh price action expected on mobile, server parts, further pressure put on AMD As revealed here earlier, Intel chose today to unleash a battery of price cuts in its high end Pentium III desktop family. (Story: Intel to announce fresh Pentium III prices) The price changes are a direct attack on AMD's recently available Athlon K7. They will be followed shortly by action in the mobile and server arenas, we are given to understand. Intel will keep its 600MHz .25micron Pentium III at the same price ($700/1000) for a little while, but has slashed the price of its 550MHz part to $490/1000. It has burnt the price of its 500MHz PIII to $255/1000, while, as reported here last month, the 450MHz Pentium III is effectively consigned to history, with a Celeron-like price of $187/1000. As reported here, the Pentium II family is now dead, with the highest specced item, the 450MHz, costing the same as the PIII/450. This pricing mechanism means that Intel has managed, in the space of only a few months, to shift its entire PII family to the PIII family. Again, as reported here, we will see Intel gradually move PIII type technology into its Celeron low-end family. (Story: Coppermine Celeron waiting in wings) The real question now is how long AMD can afford to keep its current pricing structure, or whether it can and will hold out against Intel's pricing action. In all of the marchitecture hype between Intel and AMD, the K6-III should not be forgotten. AMD wanted to pitch this part against the Pentium III, but it has not sold at all well, largely because of Intel's pricing actions throughout 1999. AMD maintains that its Athlon K7 should be compared not against the PIII family but against forthcoming Intel Will-its-release-date-be mette processors. ® See also Intel brings Pentium III launch forward (The infamous Tweedledee-Tweedledum story) Intel to send PIII/450 to gulag Pentium II family reaches end of the road Five more Intel parts to join PIII/450 in chip gulag Intel: 133MHz FSB plot to thicken sooner than thought K7 strategy is a tightrope walk AMD-Intel battle of the processors about to start Intel's PIII/600 for sale -- applies price fork to AMD K7 Do Athlon K7 figures add up for AMD?
We've now got some real feedback on this piece of Intel software, and thought it worthwhile doing a brief roundup. After having a few days to weigh the opinions of people more technical than ourselves, we are now able to give our usual unbalanced view of Intel's piece of NEWSPEED software. George Schnurer, executive editor at c't magazine in Germany, wrote us to say: "As far as I understand, the software, which I have in my lab, only enables Intel motherboards to support higher frequency processors. The only thing the prog is doing is to set up the BF-Signals for new multipliers. It definetely doesn't overtake the fixed multiplier in most of the Intel CPUs on the market. So, if someone wants to overclock a Celeron with fixed multiplier, the software won't help in any way." Over at the Anandtech forum, there is a very lengthy thread, and some of the comments are well worth reading. It's clear that the software only runs on some Intel motherboards, otherwise delivering an error message. Kyle, over at "HardOCP, said Friday he would experiment with the software over this weekend, so it's probably worth checking that site out later today. One reader said that the software is written in highly optimised assembler and could give some clues to Intel architecture if anyone wanted to take the effort to dis-assemble it. But another pointed to a line embedded in the 27K file which suggested it is, at least in part, written in C, as it contains this string: "MS Run-time Library - Copyright (c) 1992, Microsoft Corp". And it's also worth looking at our earlier stories, particularly this one, which has comments from both an Intel insider and the official Intel response. Whatever the software is, it isn't any kind of panacea, that's for sure. There is also some comment on our own message forum. ®
Far from being a quiet time of the year, August 1999 decided to keep everyone at The Register well busy. On the microprocessor front, we started the week by reporting that the Pentium II family has now reached the end of the road. Coupled to that, our favourite Japanese site Happy Cat posted a delicious Intel roadmap with some new codenames to excite the curious. We expected on the 18th of August to see a whole heap of support for the K7 chip, but little materialised, at least here in the UK. Midweek, Intel admitted it couldn't hack it in the discrete graphics chip business. Although our friends over at news.com maintained they had a scoop on this one, in reality we'd forecast that the 7x family would become dead duckies about a month earlier. Pete Sherriff caused his usual storm of protest when he released a whole set of his price/performance benchmarks, in anticipation of Celerons using Pentium III cores. (RegMark Lite shows Celeron trashing AMD Athlon and All-new RegMark99 shows Celeron outperforms Pentium III by 2.3 times and Coppermine Celeron waiting in wings) On Thursday, we learnt of a piece of Intel clocking software, NEWSPEED, which continued to attract a lot of attention through the rest of the week . Reports from Via towards the end of the week said it would go volume with the Cyrix "Gobi" S370 processor towards the end of the year. Early on Thursday, we scooped the world with our story about Alpha NT developers being laid off at a Compaq plant in the US. At a conference call on Monday, senior VP Enrico "The Cloak" Pesatori had pledged his company's support for Microsoft and also said the Alpha processor was still being pushed hard by Compaq. ®
Rambus was a busy company 20 August last. It registered a total of nine different Internet domain names, all with the word matters inside their titles. For example, it registered RAMBUSMATTERS.NET and RAMBUSMATTERS.ORG -- understandably. Less understandably, it even seemed to get caring and sharing about the opposition. So it went on to register SDRAMMATTERS.ORG, and SDRAMMATTERS.COM. Then it went further and registered DDRMATTERS.NET and DDRMATTERS.COM. Something's obviously the matter, but we noticed that its share price had managed to claw its way back to $89 by close of play Friday. None of these sites are up yet, but the company obviously has some marketing idea up its sleeve. ®