The hot topic in high performance computing these days is "GPU computing", which aims at running scientific codes on Graphics Processing Units designed for fast rendering of complex 3D graphics on high resolution displays (think video games). Claims of 100X speedups in the time-to-solution over their CPU counterparts is not uncommon and is even expected. But is it true? Can it be that much faster? What are the differences and similarities between a modern multi-core CPU and one of the latest GPUs? And how can one take advantage of such a device? Does it require an entire code rewrite or a simple recompile? This presentation will try to answer these questions and put GPU computing in perspective in today's multi-core world.