Modern computers use dynamic RAM, a technology that allows very compact bits in return for having to refresh for about 400 nanoseconds every 3-4 microseconds. But what if you couldn’t afford even such a tiny holdup? [LaurieWired] goes into excruciating detail about how to avoid this delay.But first, why do we care? It once again comes down to high-frequency trading; a couple nanoseconds of latency can be the difference between winning or losing a buy order. You likely miss all the caches and need to fetch data from the remote land of main memory. And if you get unlucky, you’ll be waiting on that price for a precious 400+ nanoseconds! [Laurie] explains all the problems faced in trying to avoid this penalty; you try to get a copy of the data on two independent refresh timers. That’s easier said than done; not only does the operating system hide the physical addresses from you, but the memory controllers themselves also scramble the addresses to the underlying RAM!For the real computer architecture nerds, there’s a lot more to it, and [Laurie] goes over it in meticulous detail in the video after the break.Thanks to [Keith Olson] for the tip!