michael@otago.ac.nz (05/20/91)
For a hint at what Apple might do next it is often a good idea to watch Motorola semiconductors (Apple are their largest customer, I think). I recently encountered some literature on the Motorola 68300 series which look suspiciously like exactly what Apple need to do a new portable. The problem with the existing Portable is that its power consumption is still too high, despite heroic efforts. Part of the reason for this is that until recently you couldn't get a 68000 that would run on less than some tens of milliamps: hence the need for heavy batteries, etc. You (still) can't get a 68020/030 that will run on less than some hundred milliamps. This is because they are dynamic devices: power consumption is proportional to speed, and Motorola say they don't guarantee an '020/'030 will function below 8Mhz. Even a CMOS 68000 only goes down to 1Mhz. So how do you do a high-performance, low-power portable? Enter the 68300 series. They are essentially a 68000 core plus peripherals on one chip. They are however capable of running at *any* clock rate down to zero(they have a new LPSTOP instruction that stops the clock) and even more interesting, they have a dynamic clock speed. They run on a 32.768kHz crystal (no, that isn't a typo). The crystal gets multiplied up by a PLL to synthesise the basic clock rate, which can be anything from 130kHz to 16.7MHz. Needless to say you have software control over the PLL multiplier and hence over power consumption. Useful to say the least. Other on-chip peripherals include dual DMA channels, fancy and very fancy timers, serial ports of various levels of complexity, and so on. None of the existing chips (68302,68331,68340,etc) would quite do for what Apple needs: the serial ports aren't Z8530 compatible and so on - but these things are designed to be assembled to a customers requirements, and Apple are certainly large enough to get one done in whatever way they like. The 68300 CPU is an interesting beast that moves halfway toward the 68020: 16-bit bus, no cache, but it has the shorter instruction timings and fast multiply and shift of the '020. It also has the loop mode found in the 68010, which altogether makes it about 1.6 times the speed of an equivalent clock-rate 68000. At 16.7 MHz its performance would be quite respectable. It doesn't have enough of the 68020 instructions that you could turn on the $MC68020 switch on you compiler and expect it to work (no bitfields for one thing). So I say unto you "Watch out for a new CPU type appearing in the interfaces", because I really can't see Apple ignoring this. Michael(tm) Hamel, Computing Services Centre, University of Otago, New Zealand ZIGONG (n.) Screeching skid made by cartoon character prior to turning round and running back in the opposite direction.