More Atari news and tips from Martin Russ.
Computer chip manufacturers are continually enhancing and upgrading their products. If they made software, then they would use version numbers in something like the usual sequence: 1.0, 1.1, 1.2, 2.0, 2.2, 3.0, 4.0... But semiconductor chips evolve in ways other than the incremental changes that happen to software. In particular, microprocessors sometimes change some of their underlying architecture and even their instruction sets. Since some sort of compatibility with previous chips is desirable, microprocessors tend to be numbered in families.
At present the two major microprocessors used in personal computers and workstations are the Intel 80n86 series and the Motorola 680n0 series. The 'n's are replaced with a number to show the relative sophistication of the chips (higher numbers also mean that the chip was released more recently). The first releases in the two families were the 8086 and the 68000. These were subsequently followed by the 80186 and 68010, and so on. What happens after the 80986 and 68090? The turn of the century!
Current IBM PCs and compatibles use the 80n86 chips — the 80386 is just reaching the end of its popularity, whilst the '486 is rapidly becoming the industry standard. Since the last three numbers are the only ones which change in a series, you often see just these digits used to describe the chip: 386 and 486, for example.
Ataris, Macs and many workstations use the 680n0 chips. STs, Mac Classics, SEs and Plusses use the ageing 68000, whilst TTs and all the Mac IIs use the popular 68030, and the Mac Quadras use the latest 68040 chip. The major leap in performance seems to happen for every other tens digit, with an approximate doubling of performance each time: so the 68020 is twice as fast as the 68000, and the 68040 is twice as fast as the 68020. In between these chips the increases are less impressive in raw power terms, but often offer other advantages. So, the '030 (last three digits again!) seems to be the chip of the moment, and you don't see many computers with 68020 processors (the original Mac II and the LC were the only ones from Apple).
Atari tend to stick with the popular chips, so the ST uses the 68000, whilst the TT and the forthcoming Falcon both use the 68030. Atari tend to be just behind Apple in releasing leading edge machines, and so with Apple releasing the second generation Classics & LCs (the Classic and LC IIs are Falcon-like low-cost 68030-based machines) and Quadras (the Quadra 950, for example), we should be expecting an Atari 68040 prototype some time soon.
So why do the TT and Falcon have a 68030 instead of a 68000? The simplest answer is pretty obvious as soon as you look at the one specification that computer manufacturers always talk about: clock speed. 68000s usually have a clock speed of 8MHz, whilst some slightly esoteric versions can run at up to 16MHz — the Mega STE, for example. In contrast, the 68030 runs at 20 or even 25MHz. Although the clock speed only gives a rough guide to the actual performance (things like disk drives, RAM caching, coprocessors and memory management can make significant differences too) you can see that three times the speed of a 68000 is going to make it significantly faster.
There is more to it than just clock speed. The 68030 can support an optional coprocessor, which can make mathematical operations much faster by removing some of the load from the 68030 itself. More importantly, the 68030 has built-in memory management (a custom chip in STs) which allows it to do clever and useful things like virtual memory, where RAM can be replaced with hard disk memory (slow, but very useful in an emergency!). The 68030 also removes one of the limitations of the basic 68000: RAM size — the 68030 can access alarmingly huge quantities of RAM (100s of megabytes if you can afford it) instead of the 4MB to which an ST is limited. The 68040 adds a floating point maths co-processor into the chip, and we may see 33MHz or even 50MHz versions soon.
But why an obsession with speed? For use with music, two things tend to make people notice when a computer is beginning to run out of processing power: slow screen updates, and time-delays on the MIDI output. Music programs tend to put the highest priority on the MIDI, which means that the screen redraws often become sluggish when the processor is working hard. Musical events tend to be clustered at bar or beat times rather than spread out in time, and this puts an uneven load on the processor, which means that as the buffering starts to fill up, so the MIDI events begin to get flammed. Add-on multi-port MIDI interfaces make things even more difficult for the processor, since there are now more MIDI buffers to cope with.
Faster and more powerful processors tend to ease the loading of the computer, which reduces the amount of queued MIDI data in the output buffers, and thus helps to speed up screen redraws. Personal computer manufacturers see things differently, of course. More power equals larger screens and more colours (colours eat up RAM and processing power alarmingly — notice the way that an ST reduces the screen size/resolution as it provides more colours). Larger screens are very useful for musical applications, as long as your favourite program will support them (typically, only the professional end of the market) and you can afford the cost of a large monitor and find somewhere to put it. Colours are rather less useful in most musical programs — a little colour can be helpful for highlighting events on a grid edit, but the trade-off in terms of performance for 256 or more colours can be expensive.
For multimedia use, the requirements are different and more in line with what the manufacturers are making. You need large screens with lots of colours, lots of RAM and enormous hard disks. Blitting big screens at more than a few frames a second really needs additional hardware, and some sophisticated picture processing techniques, such as only redrawing the parts of the screen which have changed. Adding MIDI to desktop video starts to stretch even powerful computers.
A curious side-effect of faster processors is that the computer software tends to become larger and more unwieldy as a consequence — which can mean that more processing power can result in performance only marginally better than the older version. This is particularly striking for IBM PC compatibles running Microsoft Windows 3.1 — a 50MHz 486 processor is just about adequate to run some of the latest word processors at a reasonable speed. The Mac's System 7.0 needs lots of RAM (at least 2MB), but seems to have fewer speed problems. For the ST, TT and Falcon, TOS 2.0 seems to have concentrated on consolidation of features and compatibility with existing software, rather than adding unnecessary frills, which is good news for Atari users.
A final word on comparisons. The 50MHz 486 I mentioned above may sound impressive in comparison with the 25MHz 68030 in a TT, but the catch is that a 486 needs more clock cycles to achieve the same results as a 68030 at the same clock speed. There are processors which are more efficient than the 68030 — it has been claimed that Acorn's ARM RISC processor can do more in one of its 4MHz clock cycles than a 68030 at 25MHz or a 486 at 50MHz.
Feature by Martin Russ
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!