RAM Speed

Can anyone convince me why Triple-Channel, 1066 MHz memory is better than the 667 MHz sticks in my ancient laptop? You’re not allowed to say “It’s faster” or “There’s more bandwidth,” nor will I be content with seeing technical benchmarks with no relationship to how I use a computer. I’m curious about real-world cases in which I would actually be able to notice the difference and say, “Oh yeah, that memory was slow.”

I recognize that triple-channel, 1066 MHz sticks would be “faster” than old 667 MHz ones, but I can’t recall a time when I’ve ever been bogged down by slow memory transfer. I’m usually waiting on disk, or I just don’t have enough RAM, so discussion of RAM bandwidth feels quite academic and ethereal to me.

Suppose you were given the choice between two computers at the same price: one had 8GB of RAM at 667 MHz, and the other had 6GB of triple-channel, 1333 MHz RAM. Is there a case when you would choose the faster memory over the one with more memory? Why? How much of an improvement would you see?

2 thoughts on “RAM Speed

  1. Ideally, your FSB/RAM speeds would match your CPU clock-for-clock so we could funnel data to the CPU as fast as the CPU itself was running. Since this isn’t the case, we need stuff like processor caches (L1, L2, L3) to smooth everything out.

    It’s a problem with any architecture that uses a data bus to connect disparate components.

    I believe the problem is referred to as the Von Neumann bottleneck: http://en.wikipedia.org/wiki/Von_Neumann_architecture#von_Neumann_bottleneck

  2. So to answer your question, it all depends on the processor. Generally speaking, I’d take a smaller amount of faster memory (better to keep up with the CPU) than a bigger chunk of RAM that worked more slowly.

    Who cares if you have 8GB of RAM if you can’t feed data to your processor fast enough to make use of its speed?

Leave a Reply

Your email address will not be published. Required fields are marked *