August 4, 1997

Tech View: Going Nowhere Faster By Sean Gallagher he platform strategies recently laid out by Microsoft and Intel-if you believe that they won't change by next week-have us all happily computing on Windows-based platforms at processor speeds reaching 1 GHz by the year 2000. We'll be typing "natural language" queries and commands into our PCs, and watching in awe as all that processing power churns out results in response. We'll all have more time to linger at the coffee machine. Yeah, right. Somehow, I suspect we'll all be spending our time at the coffee machine the same way we do now-waiting for our systems to catch up with us. A couple of years ago, Andy Grove got on this "free Mips and free baud" kick, as part of an effort to market Intel's ProShare teleconferencing software. The same things that have held back the sales of that product line stand squarely in the way of both Microsoft's and Intel's strategies. First, despite what Mr. Grove said back then (and whatever illusions he has now), the bauds still aren't anywhere close to being free. The network is still the bottleneck, even at 100 Mbps or 1 Gbps; the majority of the world still trudges along at 10 Mbps locally-and somewhat less outside the LAN. Somehow, I don't see that changing radically over the next two years. And as applications begin to rely more on data sources on the other side of low-bandwidth network connections-such as the Internet-that bottleneck gets narrower, not wider. Second, despite everything Microsoft and Intel would like you to believe, throwing more chip speed and more memory at applications may do wonders for application performance, but it doesn't make them scale any better. Good code and a solid architecture-neither of which are characteristics I'd assign to the world of Wintel-are what make applications scale better. Microsoft's main success over the last few revisions of its operating syst ems and applications has been wiping out any real performance benefits of the underlying hardware. Features such as the "natural language" capability planned for a future release of Microsoft Office will further drag down performance on desktops-and servers, once Windows terminals are added to the mix. Honestly, do your current Windows applications run any faster on Pentium Pro systems than their predecessors did on 486s? Sure, ease of use is valuable. And yes, the additional functionality we've gotten from the rapid bloat of Microsoft's platform has made some tasks faster and easier. But mostly, the strategy of Microsoft and Intel has been to line their wallets as thickly as possible at our expense-and their latest strategy seems to be more of the same.   Back to Labs   Send Us Your Feedback   Top of the Page

---