Previous | Next --- Slide 27 of 62
Back to Lecture Thumbnails
Firephinx

Which of these two are usually used in actual Intel chips? The first option would definitely result in quicker responses from memory since the memory controller would be able to respond with the data as soon as all of the snoop results are received. However, if grabbing data from memory does not happen as often as it does from the cache, then a lot of energy is wasted for the memory controller to grab data each time when it is probably already in the cache of another processor. How much longer latency would the second option have?