Explore BrainMass

Improving Memory Systems Reductions

Please provide a detailed step by step solution.

An embedded application requires a certain level of overall performance and after the first design phase is completed, it is evident that the chosen processing system is inadequate and needed to achieve an overall improvement in speed of at least 20%. The designers then looked at two different approaches:

(i) Improving the memory system to one with a 35% reduction in cycle time. This would double the cost of the memory system which constituted 20% of the cost of the entire system cost before this modification;

(ii) Improving the floating point ALU to one which cost 4.5 times the original and would treble the performance of the ALU. The ALU constituted 10% of the system cost before this modification.

An analysis of the application before modification shows that the ALU is used 25% of the time and the memory is used 50% of the time.

Analyze whether either of these improvements would provide the speed increase required and if both of them do, which would be the most economical to implement.


Solution Preview

Using Amdahl's Law:

Case1: Improving the memory by reducing the cycle time by 35% increases the speed by a factor of 100% - 35% or 65% (because as duty time goes down speed increases). Thus the relative speed increase achieved by adopting new memory is 1 + 0.65 = 1.65 .

As memory is used 50% of the time, we get, using Amdahl's Law, an overall speed increase as computed below.
Speed Increase = 1/[(0.5/1.65) + (0.5/1.00)]

Note that in the above expression you need to add in the factor of 0.5/1.0 to account for the other 50% of the ...

Solution Summary

The expert examines improving memory systems reductions.