graeme@labtam.labtam.oz (Graeme Gill) (01/09/91)
In article <3104@crdos1.crd.ge.COM>, davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) writes: > In article <5833@labtam.labtam.oz> scott@labtam.labtam.oz (Scott Colwell) writes: > > | Imagine what might be produced if > | the silicon technology and the effort that are put into the x86 and 860 > | groups was put into just one RISC family rather than the two RISC and one > | CISC family that Intel have today. (i.e. x86, 860, 960) > > I really don't see that any of the lines you mention are suffering > from lack of resources (maybe the 486), and no one CPU would be likely > to fit all the markets and be priced for mass marketing. > In looking at the history of different processor developments at Intel, it is interesting to read between the lines and get a whiff of the internal company politics that seems to affect the outcome more than technical considerations. "The 80960 Microprocessor Architecture" by Glenford J. Myers and David L. Budde gives some hints as to what probably went on in the development of the 80960 and 80386. Basically, the 80960 group had working full speed silicon (including on chip floating point) some time before the 80386 group. In fact the 80387 FPU core was borrowed from the 80960 FPU. At the release of the 80386, the 80960 chip was capable of providing twice the integer and floating point performance at the same clock rate, and was less sensitive to memory speed than the 80386. The 80960 was fully intended to be a UNIX style CPU engine, and included virtual memory support. What happened after that one can only guess, but the 80960K series was not released when it was ready, as it would have confused customers and sabotage sales of the soon-to-be-released 80386. In fact Intel seems to have sat on the architecturally superior 80960 for a few years, until it was "re-targeted" at embedded applications. I understand that a number of those who worked on the 960 project left in disgust. In the vacuum created by the crushing of the 960 and the architectural limitations of the 386 family, it is possible to imagine that there was the space for the the 80860 to get of the ground. Unfortunately it was yet-another architecture. The 960 project is far from being dead, but the second generation of products is being slow to emerge, and the marketing and support is strictly for embedded rather than general purpose applications. In my opinion, this is where it is a pity that Intel hasn't put more of its resources behind one family, as a hypothetical 80960CC+ (C series integer core, C series floating point, virtual memory support, + maybe graphics extensions) would be more than a match for the 80860 (or n11). If one can extrapolate from the pricing of the 80960CA, such a chip would be very well priced, and would be capable of both general purpose and embedded applications. It is really a pity that Intel had so much trouble figuring out a way of introducing one new CPU family without hurting sales of the old, that it has ended up with 3, 32 bit CPU families ! If one could re-write history, I would be tempted to re-write it in such a way that the 80960K was released instead of the 80386, with 8086 instruction set emulation in micro-code. Graeme Gill Electronic Design Engineer Labtam Australia