kilocorechip

Take a look at the future. All 1000 cores of it.

This chip, built by IBM using quite frankly old fabbing tech (32 nm), was designed by grad students at the UC Davis Department of Electrical and Computer Engineering. The impressive thing here is how each core is, more or less, an independent computer able to run its own program, with a ridiculously low energy footprint.

It’s going to take time, this chip is in no way intended to be a mass-market device, but the kind of advances it promises are incredible. With an energy efficiency in the order of 100 times better than that of a laptop, and able to scale up to 1.78 trillion instructions per second (an impressive number), it can nevertheless scale down to numbers like a still very respectable 115 billion instructions per second whilst running off of 0.7 watts. That’s low enough to run from an AA-battery whilst providing more oomph (as far as I can tell) than a modern raspberry pi.

Current modern compute architecture and programming is built around relatively few cores running very, very fast. Whilst this chip is the opposite of that (lots of cores running simpler programs) it is still very fast (when adequately powered).

It does, however, require an entirely different way of programming; new programming tools and languages, new operating systems, new programs. With what I’ve seen of current compute paradigms, however (‘the cloud’, basically), I don’t think that sort of change is either impossible nor even very far off. We’ve been moving towards smaller programs that work together on big tasks piecemeal, this is just a few steps closer.

I don’t think this sort of thing will take over everywhere – there are always going to be compute jobs that require sheer grunt horsepower – but I do believe we’re moving towards a ‘compute everywhere’ future where chips that take bites out of bigger tasks exist in everything, talking to each other constantly. I carry at least eight cores on me right now wherever I go, several more if I wear the watch I keep forgetting, and several more than that when I take both phones (don’t ask). Is it really hard to believe that we could be switching to  800 or 8000 rather than 8?

With HP’s memristor technology, this kind of ‘compute everywhere’ architecture will have come of age, and will definitely spark that next revolution that has been coming for the last few years.

We’re almost at breath-holding time, but not quite yet. Give it five years, maybe three.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s