Doubts regarding Assignment 5

Doubts regarding Assignment 5

by Jerome Shijo -
Number of replies: 2

Good evening everyone,

I’m hoping to clarify a couple of points in Assignment 5:

  1. In part 1c, when it refers to the “single-core chip with m times fewer transistors,” does that mean the cores on a 4-core design run at the same clock frequency as the smaller single-core chip?

  2. In question 2, I’m a bit confused about the Z-plot. Just to confirm: do we plot the values of P( n ) on the y-axis, and then use the corresponding n to compute E( n ), which we plot on the x-axis—so that each point represents a specific configuration? I’m not entirely sure how the points are laid out, and I’d really appreciate some clarification.

Thanks in advance!

In reply to Jerome Shijo

Re: Doubts regarding Assignment 5

by Georg Hager -
  1. The goal is to compare the m-core design (with lower clock speed) with the 1-core design (with higher clock speed). The ratio of clock speeds was calculated in Task 1b. 
  2. The Z-plot shows E( n ) vs. P( n ), i.e., P( n ) on the x-axis and E( n ) on the y-axis. Of course you can transpose it, this will not change the insight, but usually it's energy on the y-axis. The blog post to which we gave you the link shows two examples. If you change the number of cores from 1 to 52, you will get a string of data points in the diagram, each denoting a specific combination of performance and energy. The Z-plot is a very nice way of plotting this data, because straight lines denote constant energy (horizontal lines), constant performance (vertical lines), or constant product of energy and runtime (lines through the origin).