Hi, webmal-ga:
Besides the crucial.com study which you already found, I've located
some other sites which I hope will be of interest to you. For the
sake of completeness, however, allow me to quote from that first site:
"Bottom Line: If your Windows XP system has less than 256MB RAM, you
should probably upgrade. Average users will probably see the best
performance with 512MB, while those who work with large files may need
even more."
[Memory Upgrades Make a Big Difference with Windows XP]
http://www.crucial.com/uk/windowsxp/upgrades_difference.asp
* * * * * * * * * * * * * * * * * * * * * * *
This next site compares WinXP "with 256Mb, 512Mb or 768Mb of RAM" on a
variety of benchmarks:
[Windows XP - Memory Lover!]
http://www.compulink.co.uk/~davedorn/computing/windows/ramneeds.htm
"It would seem that XP has a ?sweet spot? around 512Mb where you get
the best performance for the money. Certainly a far larger amount of
RAM than Microsoft?s literature would suggest."
These tests, although done with RAM from crucial.com, appear to have
been made on an independent platform (Athlon XP 2400+, where
crucial.com reports tests from two older AMD based machines). Where
crucial.com reported only Winstone scores, increasing by 33-66% from
128Mb to 256Mb but only by 8-10% from 256Mb to 512Mb, this newer study
reports:
"What became immediately obvious was that 128Mb was simply not enough
to allow Windows XP to operate efficiently. We saw an average increase
in speed of 25% between 128Mb and 256Mb. Many people would see less
with a major processor upgrade so this was quite a startling finding.
Given that adding 128Mb of RAM is not particularly expensive compared
to a new CPU, this shows clearly that for XP users, memory may be the
first thing to check if performance needs boosting.
"The next step up was 512Mb. Here the increase was less dramatic but
still quite noticeable. We saw about a 5% increase in speed depending
on the test. Some benchmarks showed slightly higher results, some
slightly less."
Kai Chandler, the author of this study, cautions that each RAM change
counted as "a significant change for the purposes of Windows XP
Product Activation" and thus a new activation code from Microsoft each
time. This may explain why there aren't more studies being attempted!
The next two studies conclude that while the performance gains from
128Mb to 256Mb are clear, going on to 512Mb is less beneficial in
absolute terms. Their numbers (double digit improvement from 128Mb to
256Mb, at best single digit improvement above that) seem consistent
from those above. However the interpretation varies, at least in
emphasis.
This prescriptive study hosted by Rambus Developer Forum uses a
variety of benchmarks under Windows XP Pro and gives a comparison of
faster memory (DDRAM vs. RDRAM) to more memory (256Mb vs. 512 Mb),
concluding that:
[More Memory Does Not Always Mean Better Performance]
(scroll to end, cf. pg. 20 of 22)
http://www.rambus.com/rdf/rdf2002/pdf/benchmarking.pdf
"Increasing Capacity Above 256MB Has Little Impact on Many of Today?s Benchmarks"
PC World's series, Upgrades: Right on the Money, also comes out with a
256Mb recommendation for "average users". Their PC Worldbench 4
series of tests, varying memory capacities (128Mb to 1Gb), and two of
their machines running WinXP Pro produced this study:
[Memory Capacity: How much RAM is enough?]
http://www.pcworld.com/howto/article/0,aid,103769,pg,4,00.asp
"The Windows XP machines were a slightly different story. When
upgraded from 96MB to 224MB, our 1.6-GHz Athlon XP+ system from
MicronPC squeezed out 8.5 percent better performance. But once again,
upgrades beyond 256MB made little difference."
"With 128MB of RAM, if you keep opening mail or browser windows--as
most of us do these days--you'll start slogging around in virtual
memory pretty quickly, especially with a Windows XP system. At just
$20 or $30 more than 128MB costs, the added headroom of 256MB pays for
itself in productivity--so we'll we stick by our 256MB
recommendation."
Supporting numbers are provided by this chart linked at the bottom of that page:
[Next Page: RAM Upgrades: Best Value At 256MB (chart)]
http://www.pcworld.com/howto/article/0,aid,103769,pg,5,00.asp
Of course it is possible to look at benchmarks which intentionally
target memory intensive tasks, such as this one:
[SiSoft's Sandra 2004]
http://www.sisoftware.net/index.html?dir=news&location=2004_release&langx=en
But in relationship to the average user's needs such focused
benchmarks ought at most to be one among several tests to give a
rounded perspective.
A few final comments: More memory may arguably benefit the laptop
user more than a desktop user, because laptop drives (and therefore
the logistics of virtual memory) tend to be slower. My usual PC is a
laptop running Windows XP Pro and is only populated with 256 Mbytes
RAM. Naturally I don't insist to be counted an average user, but in
writing emails, surfing the net, doing light development (including a
SQL Server instance), some MS Office applications, etc. and with no
serious gaming interests at this time, I'm probably not too far off
the mark.
regards, mathtalk-ga |
Clarification of Answer by
mathtalk-ga
on
08 Feb 2004 12:06 PST
Hi, webmal-ga:
Thanks for allowing me the opportunity to clarify my answer and to
provide you with additional studies.
Here's a further performance study which deals with 128Mb vs. 256Mb vs. 512Mb:
[Best value ways to boost your PC]
http://www.apcmag.com/apc/v3.nsf/0/B20D0DF18C3F63B9CA256D44001A4D0C
"With Windows XP it?s a different story, of course. Each benchmark
rose as we stepped up to 256MB and then 512MB."
Performance on two out of three tests here improved in absolute terms
by an equal or greater amount in going from 256Mb to 512Mb than from
128Mb to 256Mb. The best percentage increase from 256Mb to 512Mb was
10% (on the 3DMark test).
I can also add this study which gives various WinXP benchmarks going
upwards from 256Mb to 2Gb:
[BAPCO SysMark 2004 White Paper]
http://www.bapco.com/techdocs/SYSmark2004WhitePaper.pdf
See specifically the chart on page 17 of 24:
"Main memory of a system is changed from 256 MB to 2 GB. The results
are shown in Table 4. Both Internet Content Creation and Office
Productivity benefit from increasing memory size. Particularly, the 2D
Creation and Communication groups are most sensitive to memory
increase. The 2D Creation group rating benefits by almost 40% when the
memory size increases from 256 MB to 2 GB. Similarly, in Office
Productivity, the Communication group rating goes up by 27% when the
memory size goes from 256 MB to 512 MB and remains constant from then
on.
"The 3D Creation group rating increases between 2.9% to 8.9% across
all four memory sizes. Over the entire spectrum, there is a big jump
in the Internet Content Creation and Office Productivity ratings when
system memory increases from 256 MB to 512 MB. However, the ratings
are almost flat when system memory increases above 512 MB."
You are certainly welcome to request a refund. However it may be
possible, since you did not mention these, that you overlooked the two
studies I originally cited besides the Rambus Developer Forum link.
For example:
[Windows XP - Memory Lover!]
http://www.compulink.co.uk/~davedorn/computing/windows/ramneeds.htm
"It would seem that XP has a ?sweet spot? around 512Mb where you get
the best performance for the money. Certainly a far larger amount of
RAM than Microsoft?s literature would suggest."
If you look at the performance gains claimed, either in the study you
found or in this one, the improvement in going from 256Mb to 512Mb is
only in the single digit percentage range. The crucial.com study
itself, which relies on a single benchmark test, the Winstone, gives
the most optimistic increase. Since crucial.com is in business to
sell RAM, this may be a case of selective reporting on their part.
Both the Rambus Developer Forum study, if read as a comparison of
256Mb vs. 512Mb DDRAM, and the PC World study both confirm the
conclusion that any performance gains in going from 256Mb to 512Mb are
of that magnitude.
Setting aside the RDF study, if you think that irrelevant, here are
the numbers from the PC Worldbench 4 study:
[Next Page: RAM Upgrades: Best Value At 256MB (chart)]
http://www.pcworld.com/howto/article/0,aid,103769,pg,5,00.asp
System/CPU: RAM type: OS 256Mb 512Mb 1024Mb
IBM Netvista PC800 WinXP 100 101 101
P4 @ 1.5GHz Pro
Micron PC
AthXP 1900+ DDR266 WinXP 115 116 117
(1.6 GHz) Pro
To my mind that study, although it makes a recommendation of 256Mbytes
of RAM as "best value", also clearly supports your claim "that
benefits begin to taper off for average users beyond 512 MB of RAM in
Windows XP".
To collate the all the results of the one link you cited and the five
different studies I'm citing, let's give the figures as percentage
improvements on cited benchmarks between 128Mb and 256Mb, then between
256Mb and 512Mb:
128Mb vs. 256Mb 256Mb vs. 512Mb
crucial.com
<your link> + 33 to 66% + 8 to 10%
Kai Chandler + 25% + 5%
RamDev Forum N/A -3 to +3%
(DDR266)
PC Worldbench + 8.5% + 1%
APC Mag + 0 to 6 % + 0 to 10%
BAPCO SysMark N/A + 15 to 17%
For the addition of memory above 512Mb, you have the PC World study
and now the BAPCO study which each report little or no performance
gains from such increases. In particular the jump from 512Mb to 1Gb
which is treated in both studies produces a gain of < 1% on the PC
Worldbench 4 results, and < 2% on the BAPCO SysMark 2004 results.
regards, mathtalk-ga
|
