Archived from groups: alt.comp.hardware.pc-homebuilt (
More info?)
john_20_28_2000@yahoo.com (jm) writes:
>Does it really matter the speed of the memory? I mean can you see a
>difference with the human eye or do I have to have a computer program
>to measure the difference? Thanks.
With your eye you can perhaps notice a 20-30% difference in speed,
if you are good at that sort of thing. The only way you are going
to consistently do better than that is to run two of them side by
side. But then you aren't really eyeballing the two speeds, you
are instead subtracting the two and eyeballing the difference.
With a stopwatch and averaging a few runs you can do substantially
better than that. But if you can't really tell the difference any
other way then how much difference does it make to you?
The 2100 versus 2700 speed is just one part of the equation. There
is also the substantial number of clock cycles it takes to begin
reading from a new part of memory. And there is the cache in your
processor that does the best it can to hide the speed of memory
from you. And the behavior of the programs you are running may
make the cache more or less effective in doing this. That doesn't
even count time used waiting for the drive, waiting for the next
net packet, waiting for the display...
Example, SDRAM versus DDR. The data supposedly pours out of the
part at twice the speed. But using a stopwatch you will usually
see 10% faster operation, at best, often as little as 5%. It is
the startup time for the blocks of reads from memory and the cache
that are contributing a much bigger part of the pie when you have
DDR, and you get 10% or 5%.
Another example: Several of the companies made lots of noise about
the increases in speed when shifting from the old PC66 to the new
and blazingly fast PC100. But those same companies didn't make any
noise about the transition to PC133. Supposedly the claim was that
the real observed increase in speed was then going to be so small
that they didn't want market backlash when most customers paid more
and couldn't see it was any faster.
Similar things happen with processor clock rates. Using the stopwatch
to measure 3.2 Ghz processors and comparing them to 2.0, you might
think you would get 60% increase in speed. But even for intensive
number crunching applications, the stopwatch shows more like 20%.
This little gem is often not mentioned in the craze to overclock
systems. How much money and time and reliablity do you want to
risk when you might get 30% increase in clock rate, and less than
half the increase in real throughput that you would need to be able
to tell the difference with your eyeballs? If that is a game and
you want to play the game, great, go for it. But maybe it would be
better to display the real actual increase in speed you get so that
others can consider whether it is worth it or not, and that is
almost never done by these folks.
Make a good informed decision, that's almost always for the best.
Facebook
Twitter
Instagram