 So at first glance, this example probably looks pretty similar to the first example. We have two machines, some CPU time, some memory access time, and one machine has a dual-core processor, the other one has a quad-core processor. You're probably expecting machine B to run the CPU instructions twice as fast as machine A does. And that would be correct. But what we have to do is be careful about what that means when we're actually looking at the total execution time. Because in this case, machine A is already experiencing some parallelization, and we'll have to take that into account. And the way we'll do that is by setting up machine A relative to a baseline that doesn't have any parallelization in it at all. So in a baseline machine, 75% of all the computation is CPU processing. But on machine A, we're already dividing that over two cores. So it's already running that in one half as much time as if we only had a single-core processor to work with. The other 25% of the computation is memory access, so that won't be affected by the dual-core processor. But I've got a one-fourth, and I've got three-eighths, which gives me five-eighths. So this is telling me that machine A is eight-fifths times faster than a single-core machine, or that it runs this problem in five-eighths as much time as it would if it had to run all the CPU instructions serially. Machine B, in turn, still has 75% of the instructions dedicated to the CPU. But now it has a quad-core processor, so it can run four of those instructions at the same time, which means that overall, the CPU instructions take one-fourth as much time to run. The other 25% of the time is dedicated to the memory, and now I get three-sixteenths plus one-fourth, which gives me seven-sixteenths. So now I'd like to compare these, so I'll convert machine A to also be in sixteenths. So now if I put these into a relative performance ratio, I can say that machine B is ten-sevenths times faster than machine A, or that machine B requires 70% of the time that machine A does to solve the same problem. So this time I have parallelization, but it's affecting both of the machines. So I have to account for that before I can calculate my actual execution time. I can't just say, oh, machine B will be twice as fast on this one-half. That would have gotten me the five-eighths, which is not the same as the ten-sevenths that I actually expect. So the final answer to the question, what is the performance of machine A relative to machine B, is that machine A is seven-tenths times as fast as machine B.