Newsgroups: comp.sys.transputer From: Graham Nice Subject: Re: Post Mortem Organization: Alex Date: Thu, 19 Feb 1998 11:00:40 +0000 Mime-Version: 1.0 Message-ID: David Boreham writes >Guys, get real about the semiconductor industry ! >Volume is everything. The vendors couldn't care less >about losing a tiny distribution account due to price. >If you knew the prices being offered to Atari, >IBM and others at the time you'd probably have >a fit. I couldn't agree more with David on the point he is making here. The price offered for any semiconductor device has very little to do with how much it costs to make but more to do with where you want to position it in the market, the value you believe it has (i.e. how much will customers pay for it) and the strategic direction the S/C company wants to take on the product line. ST is not interested in parallel processing, so why offer aggressive prices on the T4/T8's to grow the market ?. They are mature products which ST would rather not make anymore and will disappear in the next couple of years, so they are treated as cash cows - make as much money on them as you can until the end of the all time buy. If you have the potential to provide ST with volume business then you can probably command whatever price you want. The customer support required for a small low volume opportunity is not that different to the support required in getting the product designed into a high volume account. As a semiconductor company if I have the choice of trying to sell to 1000 small companies with a potential of <1000 pieces each or getting designed into a single company which has the potential of >1MU per year then the direction is clear. This is exactly what ST are doing and have significantly grown their business on volume custom designs using the ST20 core for single chip applications - into just a handful of customers. This is ST's strength - doing what a volume customer wants technically and using its volume manufacturing processes to deliver the price the volume customer wants. However, as this thread is trying to examine a business case for a new "transputer" then I thought I would share some interesting facts about 2 parallel architectures that were trying to get off the ground in 1989. One was the T9000 (which as we know is now dead) and the other was the SHARC (which is fully supported by Analog Devices and sold more units in the last 3 years than the entire transputer family ever did). In 1989 a delegation from MIT/Westinghouse were visiting Inmos to discuss the possibility of using the T9000 core as the core for the SHARC, whose architecture as defined by MIT had very similar attributes to the transputer - high processing performance, large on-chip memory and communication links. Inmos said that they were not interested in modifying the T9 core to fit the SHARC requirements (claiming the T9000 was a superior device and would be more successful !) and suggested the delegation go and talk to Analog Devices (ADI), who have some DSP cores and good processing technology. At the time one could view both the SHARC and the T9000 as potential rival devices. Where the T9 scored on its communications (VCP and associated C104 router), on-chip parallelism and targeted MIPS capability the SHARC scored on its superior MFLOP's, more and faster links and bigger on-chip memory. I'm sure some of the readers of this news group will start crying now that you cannnot make such a comparison. So be it - but this story is about commercial success, not technical wish lists. To make a microprocessor a success a semiconductor company has to bring a number of things together including: - having a sound business plan and direction for the new product signed up to by the whole corporation, - having the right process technology to support the design, - having established expertise and products in a similar product area, - having design experience to tackle the design challenge of a complex device, - using state of the art CAD tools, - treating the product as a central part of the companies product portfolio with global sales/marketing support as a thrust product, and a long term strategy to address various markets applicable to the device, - providing considerable development funding, in line with the levels required by such a development programme, into the product and future variants, - Focus, - staying power. The SHARC is a thrust, high focus, product for ADI supported by all the sales/marketing functions within ADI worldwide. It has been adequately funded to ensure its success. However the support from ST in these areas was severely lacking in most cases. As a result the SHARC has a growing market potential today and has 2nd and 3rd generations already being worked on to take the technology to >1.2GFLOPS and 4.8GOPS in 2-3 years time - and support parallel systems through link connectivity, a clear focus for ADI to maintain and grow its lead in the parallel processing marketplace. ADI have also defined low cost versions to address the single chip consumer market with a recent announcement of a $10 SHARC to be available by the end of the year. A well balanced strategy. The T9 is dead and so is any remote possibility of a future "transputer" from ST. I would admit that I am biased as I experienced considerable frustration at trying to market the transputer when at Inmos/ST (it was like trying to box with your hands tied behind your back) and now the refreshing approach and support I see (working for a Alex Computer Systems - a SHARC 3rd party board and software vendor) from ADI in supporting the SHARC and parallel processing. As far as a business model for a new transputer is concerned this is beyond the scope of news group discussion and should be a subject of a detailed evluation by an independant marketing organisation or one of the mainstream semiconductor companies. Regards and best wishes, Graham Nice (Normal disclaimers apply - views are my own etc.)