David Coveney

The astonishing power of modern computing

A photo of a Core S3 Development Kit held in the hand

Being very old (or at least, that’s how I feel being in tech!) means that after coming up to nearly forty years in technology, I’ve seen some changes.

My first computer at home, that I owned, that I could truly call my own, was a Dragon 32. It was a small, 32KB computer using the rather lovely little 6809 processor. This CPU ran at 2MHz and the system as a whole allowed me to learn a heck of a lot about computer science as a geeky teenager who was busy ignoring sports.

It didn’t exactly prepare me for my first coding job, working on an IBM MVS mainframe computer handling the payroll of 70,000 people, and distributing over £1bn a year. In fact, it prepared me terribly, because mainframes back then were similar to desktop computers today – powerful, with a ton of security features, online storage and a lot of data pipelines. But still, I knew various algorithms, knew what compilers were (though I’d only used an assembler really, but it got me half way there) and how to do a lot with a little.

That mainframe, when I started, could do about two hundred megaflops.

And today, in the picture above, I have an M5Stack on my desk. It uses the Xtensa LX7 processor. As you can see from the picture it’s absolutely tiny. And it can handle the integrated digital camera, WiFi, Bluetooth, USB… everything, really. The little CPU can run at low speeds and high speeds as appropriate to workload, and can reach over 1GHz.

Let’s put things into perspective.

The 6809 was addressing 32KB. The S3 Stack has 16MB of Flash and 8MB RAM. So 750 times the onboard storage of my Dragon 32. You can insert a 16GB microSD Card. You could put 200KB onto a C60 cassette tape back then. So that’s 80,000 times the unloadable storage! Running a lot faster too.

In terms of CPU performance, I’m struggling to find comparable benchmarks, but if we look at a 6809 you used to get 2-3 clock cycles per simple instruction – so you’d max out at around one million per second. Complex instructions would be far worse, and moving memory around was slow and expensive. The LX7 seems to get through around 0.8 instructions per clock cycle, so we can see it can max out at over a billion instructions per second. Or at least a couple of thousand times faster than the 6809.

But here’s the thing… the little computer I’m holding in my hand is at least as capable as the  contents of the data centre I started to work at in 1987.

And that, my friends, is why I feel very very old in this sector! Things have changed so much.

Yet at the same time, they haven’t. UX didn’t have a name back then, but careful thought into the design of a screen could save people a huge amount of time or reduce data errors. Saving CPU cycles then meant less of a bill (we paid by CPU time on mainframes) and could save tens of thousands a year for our team. Today, it may feel as if the CPUs are endlessly powerful, but using lower power techniques can dramatically cut down on resource use – meaning less environmental damage. I know at work we’ve taken systems running on 32GB, top spec web servers down to 512MB machines! It’s incredibly satisfying to reduce resource consumption by so much.

Let’s look at it another way. The sophisticated corporate payroll, for 70,000 people that I worked on in the eighties and nineties could easily be run today, if suitably considered, on the tiny little computer I’m holding in my hand here.

Which leads me to the real point of this piece – we as software engineers and designers have a duty to minimise resource usage. Yes, that enormous library you’ve included in your project might save you an hour of time today, but if it increases the memory and power consumption of your system so much that you need loads of cloud computing to support your system then it’s not a job well done – it’s a kludge you can hide. We have an ethical imperative to do better in systems development.

Also, low power computing is awesome and fun. Try it!

Have your say

This site uses Akismet to reduce spam. Learn how your comment data is processed.