Martin Banks, Personal Computer World 10/88 - checked

October 1988

Right, hands up anyone who knows how long a piece of string is? No, don't giggle, this is an important question for anyone in the computer business, be you maker, supplier or user. Knowing how long a piece of string is, or more importantly, YOUR piece of string, can be vital to your standing in the world.

After all, if YOUR piece of string is only a few inches long, and is made up of only a dozen or so strands, it is going to be no match for mine, which is 17,000 miles long and is made of 25,000 lengths of quarter inch diameter tempered steel. I am, by obvious definition, much more 'stringworthy' than you.

This is exceptionally good for my ego, which is (as those who know will readily testify) a weak and delicate creature that is as easily bruised as a apple falling from a tree.

It is, however, not always particularly good for my relationship with the rest of the world. For instance, having such a stupendously wonderful piece of string is of little value if what I really need to do is wrap up a rather small and insignificant parcel. Somehow, your piece of string, woefully inadequate as it is, will actually be more appropriate to life than mine.

The same analogy can be applied to a great many different aspects of life, not least of which is the personal computer. It is rather sad to see it happening, but measurement, and the associated machismo which always seems to attend it, is creeping rapidly into PC ownership.

It has long been present in the overall computer business, especially at the mainframe and minicomputer end of the scale. The most common form of measurement has been the MIPS, which stands for Millions of Instructions Per Second. This would seem to be a nice, clear way of measuring the capabilities of different computers.

Yet it isn't, and it can't be. There is also the problem of the additional consideration in deciding whether such a thing is relevant to anybody in the whole wide world except the marketing departments of computer manufacturers.

In the world of IBM mainframes or DEC minicomputers it can make a certain amount of sense to measure the performance of a computer in MIPS, for at least one is measuring like with like. In that context it is possible to get some guide between a DEC VAX 11/780 and a VAX 8600, for example, for the machines have the same architecture, the same instruction set, and the same operating system.

Even here, though, there can be problems. Specific applications can have their effect, and a VAX with much larger disk capacity or faster access disk drives may be faster in overall operation than a VAX with a notionally faster processor speed but slower peripherals.

The whole problem gets silly, however, when you start comparing unlikes. For example, how can you truly compare a Deskpro 386, which is notionally rated at 3 MIPS, with a DEC VAX 11/780 rated at 1 MIPS. The things are built completely differently, don't run the same applications and are used for different purposes. Is it of any value to the world at all?

It should be noted, by the way, that one of the touchstones of MIPSistics is that the VAX 11/780 is rated at 1 MIPS. It against this machine and rating that all other manufacturers make their claims. Yet that VAX machine, according some DEC experts, only runs at 750K Instructions Per Second (or KIPS as they are sometimes unkindly known).

One reason I mention this last bit is that I have had a letter from Neutral Ltd, complaining about the MIPS measurement situation, and suggesting that this VAXMIPS rating be expressly quoted. I don't see why I should, especially as DEC itself thinks that MIPS as a useful measurement are about as valid as that piece of string.

Neutral do, however, rightly point out the basic fallacy of MIPS. The company compares the Transputer (rated at 10 MIPS by someone) and the Motorola 68020 (rated at 3 MIPS by someone else). Yet, as the company points out, the Transputer instructions are pretty basic. It apparently takes eight instructions to load a 32-bit constant on to the register stack of the Transputer, and just one instruction on the 68020. In this task, the Motorola chip is actually 25% faster.

The nonsense then starts when Transputer fans bring forth equal and opposite arguments that show the 68020 is hopelessly slow. In practice, you see, all such arguments are definably fatuous.

They become even more fatuous when applied to personal computers, even those macho-machines used by the real whizzoes in industry and commerce. It has always amazed me that people with apparently more money than sense can convince themselves that they really need the latest, greatest, biggest and fastest box to play Flight Simulator. Now that MIPS are being bandied about the PC business as a guide to their machismo, this can only get sillier.

And yet....there is a general consensus amongst many people that there really ought to be a sensible way of getting some sort of genuine guide to how different computers compare. Price, at its crudest level, is one guide, but with so many machines available, making reliable comparisons can sometimes be difficult. And as we move inexorably beyond the PC 'standard', this task will get to be important for some.

Finding the right criteria, however, is not so easy, for not all machines run totally similar applications. Perhaps there is a role here for you, the users. Perhaps between you you can generate the tests and terms of reference which will give a reliable touchstone by which to compare totally different computer systems (not just processors).

Then again, you might prefer to go down the pub and talk about your pieces of string.

end