May 072010

I read a nice article on a German news site called about a processor function called resiliency. This function manages to correct errors that were cause by temporary or even permanent over or under powering of the processing unit. This means a resiliency CPUs can be over-clocked to a point were other processors would fail to work. On the other hand it lets CPUs run at lower power, were other CPUs would stop working as well.

Quite a crazy attempt Intel pulled off there; simply give the device more or less power than it can handle and let the device automatically correct the errors caused by it. Reminded me a bit of Jim Button and Luke the Engine Driver (original German title: Jim Knopf und Lukas der Lokomotivführer) when they made their steam locomotive fly by attaching a magnet to it.

After I read the golem article I did some digging and found a post on Intel’s blog that was written on September 23rd last year. There they also posted a video explaining what resiliency does and how it works. resiliency goes as far, as saving 22% energy at the same performance or even gaining 41% performance while using the same energy.

I’m exited to see this feature in some future peace of hardware. I’m also curious if AMD is working on something similar. What ever the future might bring, it will always be unexpected and therefore interesting.