Geek question -RAM speed/ benchmarks

[PowerUser]

Hi- I just installed new Kingston 2*1GB RAM on my Dell Inspiron 1505 (Intel
Calistoga i945PM). The earlier memory was 533MHz, while the new modules are
667 MHz (They are http://www.buy.com/retail/product.asp?sku=204884016).

The problem is that these modules are running at 533Mhz. I verified that
the 945PM does indeed support 667
(http://www.intel.com/products/chipsets/945pm/index.htm). However, these
modules are running at 533 for sure. Everest says the memory speed is
really 667, however the Memory bus 'Effective clock' is 533Mhz, with the
real clock being 267 (It doubles with dual channel).

Before I upgraded my RAM, I had taken benchmarks (Everest cache and memory
benchmarks). The new benchmarks are almost completely identical to the old
ones. This seems to be something to do with the motherboard (a BIOS
setting?). My BIOS is the newest version- A17.

Any ideas? Please!

 

[Shenan Stanley]

Hi- I just installed new Kingston 2*1GB RAM on my Dell Inspiron
1505 (Intel Calistoga i945PM). The earlier memory was 533MHz,
while the new modules are 667 MHz (They are
http://www.buy.com/retail/product.asp?sku=204884016).
The problem is that these modules are running at 533Mhz. I
verified that the 945PM does indeed support 667
(http://www.intel.com/products/chipsets/945pm/index.htm). However,
these modules are running at 533 for sure. Everest says the memory
speed is really 667, however the Memory bus 'Effective clock' is
533Mhz, with the real clock being 267 (It doubles with dual
channel).
Before I upgraded my RAM, I had taken benchmarks (Everest cache and
memory benchmarks). The new benchmarks are almost completely
identical to the old ones. This seems to be something to do with
the motherboard (a BIOS setting?). My BIOS is the newest version-
A17.
Any ideas? Please!

If you notice the difference (even if it was reporting correctly) - you need
to get out more.

Faster RAM is not going to make a noticable dent in the performance of a
machine unless you went from unnaturally slow RAM to unnaturally fast RAM -
and then - you would have changed motherbaord/processor as well - possibly
even hard disk drive and video - not by choice - but necessity.

 

[Guest]

Just curious, what is the number of CPU? T7xxx?

My point is that, as far as I know, the "true" speed you will experience is
the minimum between the Front Side Bus (FSB) and the RAM speed, isn't?

You can put 1000Ghz RAM modules, but the if the CPU FSB is only 533, nothing
will change in actual speed...

Just my two cents!

 

[PowerUser]

Perfect. That's the depressing answer (but correct) I just found out. The
FSB of <this piece of junk> processor (It's a Core Duo T2050) is 533 MHz.
The Motherboard supports 667 modules, but the RAM can only transmit data as
fast as the processor can interact with the rest of the system.

To the other poster- You *will* see a difference in benchmarks. If the
memory were running at 667, the bandwidth offered would be 10.7GB/s, or 25%
more than that achieved with 533MHz. And the question wasn't about a dent
in performance. Those were benchmark numbers. You don't argue with
numbers.

PS: How easy is it to upgrade a laptop processor? lol

 

[Shenan Stanley]

PowerUser wrote:

To the other poster- You *will* see a difference in benchmarks. If
the memory were running at 667, the bandwidth offered would be
10.7GB/s, or 25% more than that achieved with 533MHz. And the
question wasn't about a dent in performance. Those were benchmark
numbers. You don't argue with numbers.

<snip>

After years of doing this and comparing the 'benchmarks' with reality - no -
you won't *see* the difference in anything but that test - more than
likely - especially in the case you gave (the only thing you changed was
memory from 533 to 667MHz - your applications are unlikely to load any
faster or refresh to the screen any faster because it is very doubtful that
in the computer - that was your bottleneck of performance in the first
place. ;-) )

You *do* argue with numbers - because they can be played with in so many
ways to mean so many things - that the true value of them ends up being in
the eyes of the presenter - not the eyes of the beholder (if they don't
bother to argue.) *grin*

 

[Ken Blake]

PowerUser wrote:

<snip>

After years of doing this and comparing the 'benchmarks' with reality
- no - you won't *see* the difference in anything but that test -
more than likely - especially in the case you gave (the only thing you
changed
was memory from 533 to 667MHz - your applications are unlikely to
load any faster or refresh to the screen any faster because it is
very doubtful that in the computer - that was your bottleneck of
performance in the first place. ;-) )

You *do* argue with numbers - because they can be played with in so
many ways to mean so many things - that the true value of them ends
up being in the eyes of the presenter - not the eyes of the beholder

I couldn't agree more with that last paragraph. The numbers may be correct,
but deciding what they mean--how to interpret them--is very often nowhere
near obvious, *especially* if they were presented by someone who has an axe
to grind.

 

[Guest]

I do not see the point of this "debate" about the overall role of benchmarks...

The reader mentioned a very specific benchmark "Everest cache and memory
benchmarks" that to the best of my knowledge (and by its very name!) does
not attempt to measure "reality"....the reader is not asking for your wisdom
about what real-life applications might benefit from faster memory

If we were to talk about real life benchmarks, it seems obvious that the
main constraint would be HDD speed in 9 out every 10 cases; but presuming
he/she does not use memory intensive applications that might have benefited
from faster memory it rather obnoxious if you ask me...

Drod

 

[PowerUser]

Unfortunately you've completely missed the point of this post as DRod
rightly pointed out. It was about theoretical maximums not being what they
should be. Nothing to do with application or overall real world system
performance - It was about the benchmark numbers.

 

[PowerUser]

Unfortunately you've completely missed the point of this post as DRod
rightly pointed out. It was about theoretical maximums not being what they
should be. Nothing to do with application or overall real world system
performance - It was about the benchmark numbers.

 

[Bob I]

The theoretical maximum is exactly what it should be. The memory system
runs at 533. The problem is the user neglected to correctly determine
the theoretical maximum.

Unfortunately you've completely missed the point of this post as DRod
rightly pointed out. It was about theoretical maximums not being what they
should be. Nothing to do with application or overall real world system
performance - It was about the benchmark numbers.