Windows Experience Index score down-graded by doubling RAM

[churin]

How does RAM size contribute the score? The base score of my PC was set
by the Memory subscore. I changed the RAM size from 1GB to 2GB and the
momory score is down-graded from 3.4 to 2.4 and so is the base score.
The original RAM consists of two 512MB sticks with CAS latency of
2-3-2-6 and were used as single channel memory each. The new RAM is one
stick 2GB with CAS latency of 3-3-3-8.
Is there anything I gained doing this change? Did I just wasted my money?

 

[Funprice]

How does RAM size contribute the score? The base score of my PC was set
by the Memory subscore. I changed the RAM size from 1GB to 2GB and the
momory score is down-graded from 3.4 to 2.4 and so is the base score.
The original RAM consists of two 512MB sticks with CAS latency of
2-3-2-6 and were used as single channel memory each. The new RAM is one
stick 2GB with CAS latency of 3-3-3-8.
Is there anything I gained doing this change? Did I just wasted my money?

With one stick you have lost your dual-channel configuration (with
loss of memory bandwith, which is doubled by dual-channel
configuration).

 

[Tony Maynard]

Difference being is that you ran "2" sticks of 512mb which probably was
running in Dual Channel mode, so by putting in a single stick you took a hit
by cutting the channels in half..think of a 2way highway vs a 4 lane split
highway....

Not to mention due to the jump in capacity you took a few hits in the
CAS/Latency... you would be better off with 2 sticks of 1Gb

 

[Travis King]

I thought the original poster said they had it running in single-channel
mode. I would say that the latency is worse on the new RAM, which probably
caused the performance drop. I'm not running dual-channel, but have RAM
with a latency of 2.5 and am getting a 4.1 for RAM. I have 1.5GB of RAM (a
gig stick and a 512 stick). Even though I can go up to PC-3000 (not 3200),
I'm currently running them in PC-2700 as my mobo detected my PC-3000 module
as PC-2700. My other module is PC-3200, however, it appears I got my 1-gig
stick just in the nick of time a few months ago as the cost of memory is on
the rise, and that same 1-gig stick now sells for $20 more.

 

[Funprice]

I thought the original poster said they had it running in single-channel
mode. I would say that the latency is worse on the new RAM, which probably
caused the performance drop. I'm not running dual-channel, but have RAM
with a latency of 2.5 and am getting a 4.1 for RAM. I have 1.5GB of RAM (a
gig stick and a 512 stick). Even though I can go up to PC-3000 (not 3200),
I'm currently running them in PC-2700 as my mobo detected my PC-3000 module
as PC-2700. My other module is PC-3200, however, it appears I got my 1-gig
stick just in the nick of time a few months ago as the cost of memory is on
the rise, and that same 1-gig stick now sells for $20 more.

No, he said: 'were used as single channel memory each', so can still
have been dual-channel (combined).

 

[Chris]

Can I ask a question on this note ? I am an IT guy (although not hardware
so no laughing) and I have two sticks. Is it better to run one in one
channel and the other in the other channel ? (hope that made sense ??).

Another thing about the slots. I have an MSI Mobo (PT 880 Neo) and one
channel has two slots (green) and the other (purple) is only one slot. Is
there an optimum way I should have these two sticks of RAM ? I currently
have one in one of the green slots and the other in the purple.

I appreciate it in advance and hope I made myself clear ?? lol

Chris

 

[Funprice]

Can I ask a question on this note ? I am an IT guy (although not hardware
so no laughing) and I have two sticks. Is it better to run one in one
channel and the other in the other channel ? (hope that made sense ??).

Another thing about the slots. I have an MSI Mobo (PT 880 Neo) and one
channel has two slots (green) and the other (purple) is only one slot. Is
there an optimum way I should have these two sticks of RAM ? I currently
have one in one of the green slots and the other in the purple.

I appreciate it in advance and hope I made myself clear ?? lol

Chris

RTFM :LOL:

 

[churin]

Thanks everybody for your responses.
The original RAM pair is intended to be used in dual channel mode but
the pair was being used in *single* mode because of a problem if used in
dual mode.
Since they are used in single mode I thought there would be some gain in
the score by doubling the total RAM size even though the CAS latency of
the new RAM is not as fast as the original stick.

To confuse the matter I found that the score rating operation does not
result a consistant number. I switched the RAM back and forth as
Old->New->Old->New->Old, and recorded the scores as follows:
3.4 -> 2.4 -> 2.8 -> 2.6 -> 2.8
My original post was made after my seeing 2.4 with the 2GB stick.
Subsequently additional three measurements were made. I wonder how
reliable the rating score is.

Anyway, there must be a gain in performance by doubling the RAM size. It
appears that none of the subscores reflects it. Is this correct?

 

[Funprice]

Thanks everybody for your responses.
The original RAM pair is intended to be used in dual channel mode but
the pair was being used in *single* mode because of a problem if used in
dual mode.
Since they are used in single mode I thought there would be some gain in
the score by doubling the total RAM size even though the CAS latency of
the new RAM is not as fast as the original stick.

To confuse the matter I found that the score rating operation does not
result a consistant number. I switched the RAM back and forth as
Old->New->Old->New->Old, and recorded the scores as follows:
3.4 -> 2.4 -> 2.8 -> 2.6 -> 2.8
My original post was made after my seeing 2.4 with the 2GB stick.
Subsequently additional three measurements were made. I wonder how
reliable the rating score is.

Anyway, there must be a gain in performance by doubling the RAM size. It
appears that none of the subscores reflects it. Is this correct?

No, memory operations are measured, not size of memory.

 

[churin]

No, memory operations are measured, not size of memory.

Do you mean that the Memory subscore does not reflect change of RAM
size? Another word, does the memory subscore stays the same whether the
RAM size is 256MB or 2GB as long as the same RAM model is used?

 

[Funprice]

Do you mean that the Memory subscore does not reflect change of RAM
size? Another word, does the memory subscore stays the same whether the
RAM size is 256MB or 2GB as long as the same RAM model is used?

Yes, according to what I read in the description Vista itself gives
about the memory score, memory operations are measured.

 

[Barry Watzman]

I would say that contrary to your post, you had dual channel memory and
your "upgrade", while doubling the size of your memory, is operating
single channel. That cuts the memory speed IN HALF, and the benchmark
is probably based on speed, not size, once the size gets to or passes
some threshold (likely 1GB). So you screwed yourself.

Best advice: Get a 2nd 2GB module identical to the one you bought to
return to dual channel operation. However, that might be problematic,
because Windows XP 32-bit and some hardware (motherboards) don't work
well with 4GB of RAM. The best option, had you investigated this before
doing anything, would have been two modules of 1GB each.

 

[Barry Watzman]

RE: "I thought the original poster said they had it running in
single-channel mode"

That's what he said, but I believe that he didn't know what he was
talking about.

And for the same reasons, your configuration is sub-optimal as well.

On any motherboard whose chipset supports dual channel operation, you
should ONLY run the system with memory installed in matched pairs for
dual channel operation.

 

[Barry Watzman]

On any motherboard whose chipset supports dual channel operation, memory
should only be installed in matched pairs that support dual channel
operation. By definition, that means that you should never have an odd
number of modules (although having an even number doesn't automatically
mean that you can or will be running dual channel).

The modules in a pair have to be architecturally identical. That does
not mean absolutely identical (I have a system running dual channel with
on Hynix module and one Infineon module), but it just about literally
takes a degreed Electrical Engineer to figure out if two non-identical
modules are or are not sufficiently architecturally identical to do
this. So, for most people, as a practical matter, buy your modules in
pairs.

Also, they have to be installed in the proper sockets. If you have two
modules and 4 sockets, there are 4 ways to install them, and only two of
those will support dual channel operation. If you have two pairs of
dissimilar modules and 4 sockets, there are even more combinations, of
which even fewer will work.

Unfortunately, the manufacturers have done color codings of memory
sockets "both ways". That is in some cases if you have one pair of
memory modules and, say, two each red and yellow sockets (4 sockets
total), you put the two modules in both red sockets, and on other
modules you put one module into a red socket and the other module into a
yellow socket (and yes, it even matters WHICH yellow socket).

Say that there are two channels (channel A and B) and each channel has
two sockets (socket 1 and socket 2 .... so you have A1 and A2, and B1
and B2). And say that you have one pair of {suitably identical}
modules. What you want is one module in each channel (that is, one
module in channel A and one module in channel B), AND they have to be in
the same socket of the respective channels (the "1" sockets or the "2"
sockets, but both the same). Well, the problem is that sometimes the
color coded memory sockets are by channel (red = channel A, yellow =
channel B), and you would need one module in a red socket and one in a
yellow socket (and the "right" yellow socket). But other manufactures
will use the color coding for the "1" sockets and the "2" sockets.

Since it has been done both ways (by different motherboard makers on
different motherboards), you just have to sit down and look at the
manual. There is no other guaranteed shortcut if you want to be sure
that you get it right the first time.

[Undoubtedly someone is going to tell you that you have it wrong, that
your modules should both be in your green slots, or both in your purple
slots. But in fact, it depends on the make and model of your
motherboard. Like I said, it's been done both ways, unfortunately.]

 

[Richard Urban]

Your new RAM, by specification, is slower than your old RAM. It "will"
impact your score - because it is running slower.

--

Regards,

Richard Urban
Microsoft MVP Windows Shell/User
(For email, remove the obvious from my address)

Quote from George Ankner:
If you knew as much as you think you know,
You would realize that you don't know what you thought you knew!

 

[churin]

The original RAM came in pair and I spent lot of time making it work in
dual channel mode but I could not. So the sticks have been used in
*single* channel mode since then. I thought RAM size must be a factor
among others to determine the score but you indicate otherwise if the
size gets above 1GB. If that is the case then instead of getting another
2GB stick for $300(The mobo requires REG/ECC) I just forget about
upgrading RAM and will regrettably get rid of the 2GB stick at eBay.

 

[churin]

Your new RAM, by specification, is slower than your old RAM. It "will"
impact your score - because it is running slower.

Doesn't doubling the RAM size impact the score at all? Do I get better
score by changing it to one having half the original size but faster CAS
latency?

 

[Barry Watzman]

I suspect that, contrary to what you thought, you WERE operating in dual
channel mode. Your own results strongly suggest that.

I was only speculating that ram size was not a factor [above some
threshold]. Another poster says explicitly that it's not, but that
speed (which you cut in half ***IF*** you were previously operating in
dual channel mode) is the basis for the memory score measurement.

Keep in mind that the score is the score, but in real use of the
computer, it may be irrelevant: Doubling memory size from 1GB to 2GB
might not increase your score, but it may very well increase the actual
speed of the computer in SOME applications. On the other hand, once you
get to 1GB, for many applications, additional memory doesn't do much.
Similarly, for some applications, the difference between single and dual
channel operation is nearly double, while for other applications, it's
nearly zero.

Still, the "right" way to deal with memory on any system that supports
dual channel operation is to install memory in matched pairs so that
dual channel operation is enabled. You never gave us the exact
motherboard you are using, so I can't say with absolute certainty if
your system does or does not support dual channel memory. However,
empirically, your results suggest strong that whatever you believed, you
were in fact running dual channel before and are now running single channel.

 

[Richard Urban]

Stop worrying so much about the score. You computer runs best when in dual
channel mode. Get two "matched" sticks of RAM and do such.

There are companies that sell Matched RAM specifically for this purpose. If
you had been using this RAM you would not have had the problems you did in
trying to run in dual mode.

The score is based upon the "speed" of the operating system, as well as
certain minimums. After you pass that minimum the speed of the electronic
devices are paramount, I would suspect.

--

Regards,

Richard Urban
Microsoft MVP Windows Shell/User
(For email, remove the obvious from my address)

Quote from George Ankner:
If you knew as much as you think you know,
You would realize that you don't know what you thought you knew!

 

[churin]

I assembled the PC and installed the original RAM pair in a specific way
so that it operates in single channel mode. Memory test clealy indicates
"Single Channel (64bits)".

Let me report results of my further experiments:
Originally the Memory subscore was 3.4 with 1GB(CAS 2-3-2-6) which
prompted me to replace 2GB(CAS 3-3-3-8) and I did. Since then I switched
back and forth and determined that repeatable measurement results are
3.0 with 1GB(512x2) and 2.6 with 2GB single stick. Then I found that the
Memory subscore can be measured to be 4.4 either with the pair RAM or
the single RAM.
The following scores ane with ECC enabled, ECC disabled, and ECC enabled
again:

512MBx2 : 3.0 --> 4.4 --> 4.4
1GB Single: 2.6 --> 4.4 --> 4.4

Note that after each change the score was updated. I thought increase
of the score after ECC was disabled made sense but it did not revert to
the previous score after ECC was enabled again. I thought the score of
4.4 looked faulty so that I did the same experiment wiht single 512MB
RAM. the resutls are as follows:

512MB Single: 2.6 --> 2.9 --> 3.0

The pattern is now different. I am not sure if the 4.4 is reliable score
or not but I am inclined to believe I am better off using the single 2GB
stick.

I would appreciate it if the participants in this thread would post your
scores with your hardware configuration. Mine are as follows:

Hardware:
Mobo: K8T Master2 FAR
CPU: Opteron 244(1.8MHz, dual)
RAM: Patriot Single 2GB PC3200 REG/ECC CAS:3-3-3-8
HD: PATAx2, SATAx2, WVT is on 20GB partition in a PATA
Display Adapter: Radeon X700

Scores:
Processor: 4.7
Memory(RAM): 4.4
Graphics: 4.9
Gaming graphics: 4.6
Primary hard disk: 5.2

I suspect that, contrary to what you thought, you WERE operating in dual
channel mode. Your own results strongly suggest that.

I was only speculating that ram size was not a factor [above some
threshold]. Another poster says explicitly that it's not, but that
speed (which you cut in half ***IF*** you were previously operating in
dual channel mode) is the basis for the memory score measurement.

Keep in mind that the score is the score, but in real use of the
computer, it may be irrelevant: Doubling memory size from 1GB to 2GB
might not increase your score, but it may very well increase the actual
speed of the computer in SOME applications. On the other hand, once you
get to 1GB, for many applications, additional memory doesn't do much.
Similarly, for some applications, the difference between single and dual
channel operation is nearly double, while for other applications, it's
nearly zero.

Still, the "right" way to deal with memory on any system that supports
dual channel operation is to install memory in matched pairs so that
dual channel operation is enabled. You never gave us the exact
motherboard you are using, so I can't say with absolute certainty if
your system does or does not support dual channel memory. However,
empirically, your results suggest strong that whatever you believed, you
were in fact running dual channel before and are now running single
channel.

The original RAM came in pair and I spent lot of time making it work
in dual channel mode but I could not. So the sticks have been used in
*single* channel mode since then. I thought RAM size must be a factor
among others to determine the score but you indicate otherwise if the
size gets above 1GB. If that is the case then instead of getting
another 2GB stick for $300(The mobo requires REG/ECC) I just forget
about upgrading RAM and will regrettably get rid of the 2GB stick at
eBay.