# CPU Memory Hierarchy: Calculating Average Memory Access Time

(From Schuam's Outlines Computer Architecture, 2002, page 193, problem 8.7(b))

Suppose I have the following memory hierarchy of:

CPU <-> SRAM <=> DRAM <=> DISK

SRAM has 5 ns access time

DRAM has 60 ns access time

DISK has 7 ms access time.

If the hit rate at each level of memory hierarchy is 80% (Except the last level of DISK which is 100% hit rate), what is the average memory access time from the CPU?

So I start the problem... here are my calculations:

For the DRAM Level the access time is:

$$T_{DRAM} = (0.8)(60 ns) + (0.2)(7 ms)$$

$$T_{DRAM} = 1.448 \mu seconds$$

For the SRAM/CPU Level the access time is:

$$T_{SRAM} = (0.8)(5 ns) + (0.2)(1.448 \mu s)$$

$$T_{SRAM} = 293.6 ns$$

Now for the problem, the solution manual for the book says the answer is:

$$T_{SRAM} = (0.80)(5 ns) + (0.20)(0.80)(60 ns) + (0.20)(7 ms)$$

which I calculate to be: 1.4136E-6 seconds and they calculate to be:

$$T_{SRAM} = 280,0136.6 ns$$

My answer is "293.6 ns", and the book's solution to the problem is "280,0136.6 ns"

Who is right and why?

• How many ns are in a ms? ;) Feb 5, 2019 at 23:28
• $$1 ms = 10^{-3} sec$$ $$1 \mu s = 10^{-6} sec$$ $$1 ns=10^{-9} sec$$ Feb 5, 2019 at 23:31

Your formulas look fine to me, but there's an error in the numbers: $$T_{DRAM} = (0.8)(60 ns) + (0.2)(7 ms) \\ T_{DRAM} = 1.448 \mu s$$ This is incorrect, with $$\0.2 * 7\$$ ms in the sum, the answer should be at least 1.4 ms, not µs.

Note that 1 ms = 1000 µs = 1 000 000 ns.

Compare:

$$T_{DRAM} = (0.8)(60 ns) + (0.2)(7000 ns) \\ T_{SRAM} = (0.8)(5 ns) + (0.2)(T_{DRAM}) \\ T_{SRAM} = 293.6 ns$$

with:

$$T_{DRAM} = (0.8)(60 ns) + (0.2)(7000000 ns) \\ T_{SRAM} = (0.8)(5 ns) + (0.2)(T_{DRAM}) \\ T_{SRAM} = 280013.6 ns$$

If you substitute $$\T_{DRAM}\$$ in the $$\T_{SRAM}\$$ expression, you get:

$$T_{SRAM} = (0.8)(5 ns) + (0.2)(0.8)(60 ns) + (0.2)(0.2)(7000000 ns)$$

The expression you quoted from the book is slightly different (only one $$\(0.2)\$$ in the last term of the sum). That looks like either an error in the book, or in your quote.

• ahh...ok... i see the error... was using 10^-6 instead of 10^-3... thanks for confirm that books equation is wrong... but strangely their answer was right in the end... Feb 6, 2019 at 0:10
• Workings/formulas get copied and pasted. You go back and edit contents. Sometimes you fix all, sometimes you do not. Usually you get the final answer correct. Feb 6, 2019 at 1:06
• @MrCasuality Check my update; if the book actually has $(0.2)(0.2)$ in the last term of the sum, the equation in the book is also correct! Feb 6, 2019 at 15:43
• typographical error in book... that's what was confusing me. Feb 8, 2019 at 13:33