If you’ve been following technology for long enough then you probably remember how processor speeds kept going up at a remarkable pace during the nineties and early naughts. You’d buy a Pentium with a 166 MHz CPU only to find it to be obsolete a few years later when CPUs were already reaching GHz speeds.
But since then, speed improvements have been fairly limited. It’s rare to find processors operating above 4 GHz in consumer devices, due to the physical limitations of the space and materials that we’re working with. That’s why, during the past decade, researchers and engineers have focused on multi-core processors instead of trying to push the clock speed higher up.
Multi-core processors are easy to understand: instead of one really fast core you have two, or four, or ten million slower cores that work together. So if you can perform a task on a single core you can do it twice as fast on a double-core processor, right? Not exactly. Parallel processing has a number of issues that lower its efficiency by a considerable margin.
Enter MIT’s Swarm. The newly-developed chip architecture promises to considerably speed up multi-core and parallel processing. It also promises to make a developer’s life much easier, because Swarm can handle many of the issues that make parallel processing inefficient directly through its hardware.
For example, when using multiple cores to process a task, one core might need to access a piece of data that’s being used by another core. Developers usually need to write code to avoid these types of conflict, and direct how each part of the task should be processed and split up between the processor’s cores. This almost never gets done with normal consumer software, hence the reason why Crysis isn’t running better on your new 10-core Intel. Meanwhile, when such optimization does get done, mainly for industrial, scientific and research computers, it takes lots of effort on the developer’s side and efficiency gains may sometimes still be minimal.
But Swarm handles all of this, mainly through its hardware architecture and some customizable profiles that can be written by developers in a fraction of the time needed for regular multi-core silicon. The MIT researchers tested out a 64-core version of Swarm, against some highly-optimized parallel processing algorithms. Swarm came out on top, offering three to 18 times faster processing. In its most impressive result, Swarm achieved results 75 times better than the regular chips, because that particular algorithm had failed to be parallelized on classic multi-core processors.
That’s highly impressive, and becomes even more so, when considering that Swarm might also offer highly improved power efficiency – a huge problem in today’s consumer and industrial systems. Researchers are now working to improve these results further by experimenting with the hardware and programming models.
There’s no clear indication this technology will make it into our consumer devices anytime soon, but we can only hope.