Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!elroy.jpl.nasa.gov!decwrl!ogicse!milton!sumax!ole!ray From: ray@ole.UUCP (Ray Berry) Newsgroups: comp.sys.ibm.pc.hardware Subject: Re: Math chip wars... Keywords: 80x87 Message-ID: <2041@ole.UUCP> Date: 26 Jun 91 04:36:04 GMT References: <1991Jun20.002949.4128@mlb.semi.harris.com> <5748@dftsrv.gsfc.nasa.gov> <1991Jun24.174305.249@gtephx.UUCP> <1991Jun25.112851.912@mlb.semi.harris.com> Organization: Seattle Silicon Corp., Bellevue, WA. Lines: 52 routh@eltanin.rtp.semi.harris.com (Kevin Routh x622) writes: ... >No bashing was intended, I'm sorry it was perceived as such. It was a >combination of amazement and envy. I'm amazed that it's possible to >drop the prices that much (the costs must have dropped or the margins >are huge). And as someone in the chip industry, I envy their status. To illustrate just how flexible Intel can be on pricing, listen to this. A few weeks ago I attended an Intel dog/pony show on the 960 series 32 bit cpus intended for embedded systems applications. Basically this fam- ily consists of 5 parts, which are analogous to the 386SX and 386DX, plus 2 parts that would be like an SX/DX but with a built-in FPU, and lastly a high-end part with built-in FPU (but very fast). Ok. These are all 32 bit high-performance CPUs. Lately Intel has been running ads for the low-end member (the 960SA) which has a 16 bit bus interface and no FPU. They're offering these at $20 at 1K units. Through the course of the presentation, the speaker explained to us that the FPU's in the 960 devices were essentially the same as the 387. At the end of the presentation I was curious about the 1K pricing for the 960SB (16 bit bus interface, but *with* the FPU on board- so I asked the price. The reply? $26.00. In other words, you pay another six bucks for the FPU. Another telling tidbit was a viewfoil that showed the five 960 family members positioned on a cost/performance graph along with the 386/486 parts. According to this particular viewfoil, the high-end 960CA was shown rating *above* the 486 in performance, while it was floating around the $100 mark in price. Never one to miss an opportunity I inquired as to the explanation for this mysterious situation - was the 486 *really* that poor a design? There was some hemming and hawing and mumbling about a "different market focus"... and a few snickers around the room. Now Intel is successful and you can't argue with success. And they spend scrillions on R&D and guess where it's coming from- the X86 family. Actually I have no problem with that. Or else we might all be learning Japanese someday to figure out how to read the data sheets for our FPUs :-) OTHO, Intel has been paying back FPU development costs for a *long* time - 10 years? Motorola has been selling FPUs around the $1-200 for a long time and I doubt they're losing money doing it. Cyrix, IIT, etc., saw a good thing and decided to try and cash in on it. They knew it wouldn't last forever. At the Intel pony show, I *did* miss one good shot. The speaker was soliciting replies from the audience on things they looked for when selecting a CPU for a high-performance embedded systems application. The obvious answer somehow escaped me: "Second source from AMD" :-) -- Ray Berry kb7ht uucp: ...sumax!ole!ray CIS: 73407,3152 /* "inquire within" */