Reader Eric Pobirs writes: There are a number of flaws in The Register analysis of Intel's future roadmap for basic PCs. Audio and DVD will not be a problem. I've observed SoftDVD playback in Windows of amazing quality on 350MHz PII systems that did not have motion compensation on their video cards. Motion compensation is rapidly becoming a generic feature in video chips that will quickly migrate into the sub-$1000 sector and can reduce CPU overhead for DVD playback by up to 30 per cent -- 40 per cent has been claimed but not observed by anyone I know. Coupled with the eventual appearance of the Katmai instructions to the Celeron line performing DVD playback will not overwhelm the system. Piling on the entirety of digital audio processing (AMR picks up the analogue end if the output isn't going to digital speakers) won't add that much to the load. If the OS can be made to give priority to the constant throughput needed by the video playback going to a TV the system should still be able to perform more 'bursty' tasks like Web surfing while someone enjoys a movie. You have to keep in mind who these machine are intended for. The kind of work described above is what the consumer thinks of when he hears the term 'multitasking' (assuming (s)he knows what that means -- Ed), not how many IP connections he can maintain. When an Intel guy says the soft AV isn't there yet he means: 'We can't offer a 500MHz Celeron with KNI and 512K on-die cache for a sub-$1K PC design yet.' Two years from now it won't a big deal. First, KNI has to earn its keep as a premium feature. This brings up another misunderstanding. Intel doesn't make a 45 per cent markup across the line. It make almost nothing on some of the low-end products. That isn't their purpose. They exist to keep AMD, Cyrix and the rest from becoming profitable enough to eat into the high-end where Intel does command massive margins. Without substantial profits (or any profits, really) the cloners can't ever hope to fund the R&D and facility contraction needed to compete where the serious money is paid for systems. Intel would happily take a loss on low-end chips to keep the cloners at bay if there weren't legal strictures against that. If the cloners give up, prices might increase a bit but Intel would still forgo low-end margins for the sake of market growth. Today's Celeron buyer might be tomorrow Merced user. Not that deep pockets alone would do the trick. IBM Microelectronics has gotten tons of press for their copper and SOI announcements but delivery has been less than impressive. So far it has only managed to demo 400MHz PPC chips just in time for that to become the secondary speed in the Intel world. Actual production seems to be another story, forcing Apple to push iMac at tradeshows like Seybold that normally wouldn't give the time of day to such a low-end machine. I sometime wonder if the ancient x86 architecture is as much help as hindrance to Intel's success. If it weren't bound to that instruction set and could start from scratch like the competition they may never have had the motivation to pull so many rabbits from hats. ® Click for more stories