0
\$\begingroup\$

I see a frequency in RAM specs which is different than the frequency of the system clock. Having a separate clock makes sense if RAM can't run at the same frequency, but is there a constraint that its frequency must be an integer divisor of the system clock's frequency so they can interface properly? The two frequencies on specs are often not multiples of each other but perhaps the RAM is under clocked until it is.

There are also future plans to clock RAM faster than the system clock. How is this possible--does the CPU ask for multiple memory instructions per cycle?

\$\endgroup\$
2
  • 1
    \$\begingroup\$ If those are your only questios, how they relate to each other depends on what system you are looking at, and whether or not they must relate by an integer factor also depends on the system. We don't know what system your question is about and what RAM it has, so it can't be answered. \$\endgroup\$
    – Justme
    Nov 24 at 14:11
  • \$\begingroup\$ Not integer divisor (since in some cases the memory frequency will be higher than the CPU frequency), but the ratio of the two frequencies is usually a rational number since it vastly simplifies clocking if they're synchronized like that. \$\endgroup\$ Nov 27 at 2:32

1 Answer 1

1
\$\begingroup\$

The DRAM interface doesn’t have to share a clock with the CPU.

But there is some benefit to carefully choosing the relationship between the two clocks: it influences the clock boundary crossing and thus the CPU-DRAM latency. Some ratios will yield higher performance even while trading off the max clock rate of each.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.