Newsgroups: comp.sys.transputer From: "Stephen Maudsley" Subject: Re: Post Mortem Organization: esgem limited Date: Tue, 17 Feb 1998 09:12:49 -0000 Mime-Version: 1.0 Content-Type: text/plain; harset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-ID: <6cbv92$dfh$1@uranium.btinternet.com> Andy Rabagliati wrote in message <6ca7o0$31c$1@news1.rmi.net>... >According to Magnus Paterson : > >> NewsMan wrote: > >> > I wonder though if anyone has performed a detailed business >> > analysis of why the product failed. >> >> The T seemed to be a solution that hadn't properly identified the >> problem it was trying to address. I built a couple of hundred >> transputers into high-performance astronomical instruments, but >> all interfaces had to be built in-house or produced by third-party >> companies such as Sunnyside and Sundance. > >I think the T4 and T8 were fine, apart from a disastrous effort to >ignore C as a programming language. But that was fixed, first by >Kirk Bailey, then by others, including Inmos. > >The problem was pathetic followup - nobody even thought of designing a >successor until much too late. > The T800 was introduced about 1986/87 and the T805 87/88 - ten years ago. Moore's law says that you can get twice as many transistors for a given area/cost every 18 months which seems to give a rough metric for microprocessor development. So if you accept that the T800 was state-of-the-art on introduction you would now need an equivalent to a T800-1300 to be competative (which is roughly where Sun, Intel, HP, IBM seem to be with their fastest devices). The T9000 would have got back onto the curve if it had been available when it was launched in San Jose in '92. '92 must have been the point where other processors had enough cost/performance benefits to overcome the advantages of integrated parallel processing - AFAIR that was when Parsytec began their "occam on the PowerPC" project and Texas DSPs were doing quite well. I agree with Andy that much of the early software support problems were in hand but there were no new processors to keep momentum in the commercial activities. --------------------- Stephen Maudsley Stephen.Maudsley@btinternet.com esgem limited Tel/Fax: +44-1453-521626 Mobile: +44-370-810991 ---------------------