7 both sides must remain equal we can solve directly

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: remain equal.) We can solve directly for k without knowing the other terms: So, no matter what the clock rate, instruction count, or CPI, you would need to increase the clock rate to 171% to decrease time by 30% if you pay a 20% CPI penalty. The answers become: Answer for follow-up question: Different processors may have different instructions. If a processor has more powerful instructions, it is likely to have a slower clock rate. (Some processors get around this by internally converting complex instructions into simpler ones.) On the other hand, a processor with more powerful instructions may be able to finish a task in a small number of instructions, where a simpler processor may require more instructions for the same task. So, when comparing processors, a processor with a slower cycle time would finish a task more quickly if it has more powerful instructions and requires fewer instructions for the same task. 2. Exercise 1.11: Each part of this problem relies on the formulas for the cost of an integrated circuit (from lecture and page 46) and simple formulas for computing area. Note that the book formulas approximate the actual formulas - we are only interested in this approximation for this problem. 1.11.1: First, compute the area of each wafer: Next, compute area of each die: Next, calculate the yield using the formula from page 46: ( ⁄) ( ⁄) 1.11.2: To find the cost per die, use the formula from page 46: 1.11.3: This is a repeat of part #1, except that the original data is scaled slightly. This approximates what would happen as feature size gets smaller: Compute area of each die assuming there are 10% more dies: Next, calculate the yield using the formula from page 46, except assume defects have increased by 15%: ( ⁄) ( ⁄) So, even though the defect rate went up significantly, the decrease in feature size compensated for it (in this example). Answer for follow-up question: A very large die is very likely to include a defect. Consider the smaller defect rate from the problem: 0.018 defects/cm2 on averag...
View Full Document

This note was uploaded on 01/16/2014 for the course ECE 3810 taught by Professor Peter during the Fall '11 term at Utah.

Ask a homework question - tutors are online