"James Waldby" wrote in message
...
...
In other words, different sizes work together ok, but may run slower
because dual-channel access won't work with different sizes lined up.
(See http://en.wikipedia.org/wiki/Multi-channel_memory_architecture
picture of two yellow and two orange slots, one of each per channel.
You didn't say if you have Minitower D530, which has 2 channels and
4 DIMM slots, or Ultraslim Desktop D530, which has 1 channel and 2
DIMM slots. With only one channel, "dual-channel performance" isn't
possible anyway.) If you are running low-end apps on small data sets
on an obsolete computer, dual-channel performance probably won't matter.
...
jiw

Here's what's happening behind the curtain:
http://en.wikipedia.org/wiki/CPU_cache

When the CPU doesn't find the memory address it wants in its built-in
superfast cache memory it loads in a new block of addresses from the RAM
sticks. I think the double, triple, quad data widths speed up the block
transfers rather than access to a single random location, so the perceived
memory speed really depends on subtleties of the hardware and software.

A compiled integer FOR loop runs at half the CPU clock speed, much faster
than system RAM access time would permit, showing that it executes from the
instruction and data caches. No one has paid me to write timing tests that
trigger cache misses.

For the user the PC either is or isn't fast enough and the only real choice
is to replace it. I sometimes hit that limit recording HDTV on a 2.2GHz
machine with a too-small L1 cache. Its 3GHz replacement loafs along at 30%
CPU useage.

In simple terms the cache contains the tools and parts you have out on your
small workbench, RAM is those in the bigger cabinet and the hard drive holds
the stuff in the huge warehouse.

jsw