The RAM "limitation" for WinXP

[oops]

First a while ago I heard someone said he saw a web site posted testing resault of a WinXP system under different amount of RAM.

And the test shows once the WinXP has more than 768MB RAM, the performance will decrease.

That's just "heard about".

-------------------------------------------------------------------------------

BUT, just now, I did a test myself !!!
(I am on WinXP Pro)

I ran my 2.4B at default speed with 3:4 ratio for RAM.

First I ran test with 2 sticks of 512MB Corsair DDR PC3200 RAM, that's total 1024MB PC3200 RAM.

Then I took out 1 stick, and I ran the test with 1 stick 512MB.

and PCmark2002 shows

(under exectly same condition)

with 1024MB RAM

CPU 5796
Mem 5770
HDD 925

---------------------
with 512MB RAM

CPU 5821
Mem 5972
HDD 950


That was under exectly same condition.
here are the links to my resaults
(NOTE: PCmark shows CPU frequence 50MHZ BELOW the actually speed)

1024MB RAM
http://service.madonion.com/compare?pcm=843515

512MB RAM
http://service.madonion.com/compare?pcm=843591


What do you think now??

So many people said "WinXP" can use unlimited RAM. Worth it?

 

[looktall]

xp can't use unlimited ram.

i believe it's limit is about 4gb. maybe it was 8gb? i forget, but i do recall reading what the limit was, somewhere awhile back.

as for more ram decreasing performance, i don't know about that, but i do know unless you have some heavy duty, high ram using apps, anything more than 512mb is just going to be wasted.
perhaps that's why you got a different PC mark score?

or perhaps the difference is due to your testing methods?
were both tests done after a fresh reboot? fresh installs, etc?

 

[oops]

or perhaps the difference is due to your testing methods?
were both tests done after a fresh reboot? fresh installs, etc?

lol,

Did you read my entire post?


Ok, here is the very clear "steps",

First I normally use 512MB RAM,

that's why I have to "turned off" the PC to add another stick of PC3200 in, and then boot into windows, then test the PCmark with 1024MB RAM.

Then turned off the PC, then took out one stick. Then turned on the PC, then ran the PCmark again with 512MB RAM.


Clear now?

 

[Stedeman]

was the new stick cl/cas 3 and the old one cl/cas2!?! if so when you added the new stick your system will move to the cl/cas 3 this will slow you down unless you where going over the first 512

 

[oops]

was the new stick cl/cas 3 and the old one cl/cas2!?! if so when you added the new stick your system will move to the cl/cas 3 this will slow you down unless you where going over the first 512

Both are cas 2.5, but I set 2-5-3-3 in BIOS. (but both can run 2-5-3-3 at 200MHZx2 and above)

 

[Stedeman]

just to be 100% sure try switching the sticks and running the test another time

 

[oops]

just to be 100% sure try switching the sticks and running the test another time

Both tests were under exectly same condition, all the setting in BIOS are exectly the same(didn't change it).

The only difference is the first time was 2 x 512MB Corsair PC3200, the second time was just 1 stick of 512MB Corsair PC3200.

I had to turn off the power to put the extra stick in and take it out, so both tests were right after fresh boot.

I also tested with 3dmark2001SE, the difference not as much as PCmark......about 20 points difference(out of 12k points, video card at default), but that's video stuff. So I think basicly overkill amount of RAM hurt CPU and memory performance, maybe hard drive too.


By the way, the other stick of RAM was from mine the other system(I wouldn't want to have overkill amount of RAM in one system)

So 512MB is enough, don't waste money to hurt the performance

 

[oops]

just to be 100% sure try switching the sticks and running the test another time

Well, if you really want then I can do it later......didn't sleep all night......

 

[ARGON]

First off, thanks for testing that! For the serious overclocker where every point on the bench counts, this is useful. What you might want to try also is changing the banks that you are using. On most boards merely changing the banks that the modules are in causes a quite noticible difference. I would suspect that this is what is going on here. Just a thought though.

For those using this information to make buying decisions, remember that more RAM is more RAM. A few marks lower in the benches doesn't mean that RAM is being wasted necissarily.

 

[oops]

PCmark2002 is pretty complex, it runs over 30 different tests, so I think it's fair.

Also, 1 x 512MB Corsair PC3200 is around $200($180 to $220), think before you order another one add into your system----"does the extra $200 worth the performance change?"

Just my 50 cents

 

[JaY_III]

less RAM give faster seek times.
open photoshop or any other program that sucks RAM up and do your bench again.
I would bet 1024 will win.

For example:
a Celeron with 0KB L2 cahce was and is faster at acessing RAM than a Celeron (same speed, 2nd place) with 128 L2 and P2 (same speed, 3rd place) with 512 L2.

I'd love to have 1024MB of RAM over 512MB.....
As i dont know the last time I only had 1 program open.

As for how much RAM windows can use:
9x :< or = 512MB
2K : < or = 4 GB
2K server/adv server/datacenter < or = 32GB
but do note 32bit CPU's can only address 4GB of RAM without mapping

 

[S_Wilson]

One thing you guys haven't touched on is that most modern high performance motherboards will adust timings when multiple sticks are in use. Probably for stability issues. So if tested with 1x512Mb compared to 2x256Mb you would probably get similar results comparing 1x512 vs. 2x512. I don't think the issue is with windows as much as the hardware.

I think it says in the overclockers bible that fewer sticks is better. It was mainly because of stability and trying to get high fsb with pc66 and pc100 ram. With todays mainboards and high performance ram, it is the small performance hit you will take because the makers have built in some safeguards to insure stability with multiple sticks.

 

[rottenspam]

The other thing I would consider too, is that the score difference are probably within the margin of error of the bench mark. As in if you ran it a couple of times it would show that much variation even if you didn't change the RAM. I don't think this demonstraits anything other than 3Dmark dosen't take advantage of larger amounts of RAM. The differences are less than 1% and aren't even perceptiable.

 

[oops]

The other thing I would consider too, is that the score difference are probably within the margin of error of the bench mark. As in if you ran it a couple of times it would show that much variation even if you didn't change the RAM. I don't think this demonstraits anything other than 3Dmark dosen't take advantage of larger amounts of RAM. The differences are less than 1% and aren't even perceptiable.

I ran both PCmark and 3dmark.

The difference in 3dmark was min.
but in PCmark the memory score difference was about 4%.

Yes, I agree the score is a little bit different every time you run it, but no big difference.

My point is, does it worth extra $200 to buy something gives a close resault like that?

 

[su root]

Well, I heard somewhere that 384Mb ram gives the best performance.

When you install more than one stick of ram, the seek time will go up, which slightly slows down your system.

The big advantage of more ram is when the ram fills up, and it needs to page to disk (which takes gobs of time)...

 

[zachj]

I agree that the test would probably be different if 1 stick were used instead of two, but probably not by any appreciable difference. Also, did you only run the benches once, or did you run them a few times and average the scores, excluding any particular high and low scores? That is the best indicator of actual performance, as it gives the average of many tests. If three houses on a street are 200,000 dollars and one is 3,000,000 dollars, the mathematical average of those four houses is 900,000 dollars, but that's not an accurate representation of the facts. The average home on that street is 200,000 dollars because that multi-million dollar house is an unaverage home that scews your results. So if you wanted to/had the time to run the tests again, doing it a few times under each situation, and then averaged the scores, I would take your results a little more seriously. (not calling them bunk, just saying . . .grain of salt)

Z

 

[JaY_III]

My point is, does it worth extra $200 to buy something gives a close resault like that?

If you sit around and run becnhmarks all day, NO
If you run multiple programs and use your Computer, YES

but hey, that just my $.02

 

[oops]

If you sit around and run becnhmarks all day, NO
If you run multiple programs and use your Computer, YES

but hey, that just my $.02

Those 30+ tests in PCmark included most daily use, and how many of you would watch DVD, play video games, listen to music at the same time?? Even you do, that 1GB RAM won't help much. The term of "multiple tasks" to most of you probably ------checking email, with several more brower window open and play music at the background, that's all. Most of you will never run another program in the background when watch DVD or play 3D games(because you know that's going to hurt the performance too much) Does that really require 1GB of RAM?

 

[ARGON]

Originally posted by oops

Those 30+ tests in PCmark included most daily use, and how many of you would watch DVD, play video games, listen to music at the same time?? Even you do, that 1GB RAM won't help much. The term of "multiple tasks" to most of you probably ------checking email, with several more brower window open and play music at the background, that's all. Most of you will never run another program in the background when watch DVD or play 3D games(because you know that's going to hurt the performance too much) Does that really require 1GB of RAM?

You don't have to be doing all of that at the same time to fully utilize your RAM. Once you open up any application, it is cached. If you close it and then re open it, you won't be reading any information of the HDD.

During a typical Windows session for me, my entire OS is cached, along with ACAD2000, UT2003, TacOps, IE, Word, Excel, Access, Outlook, Matlab, Mathematica, Windows Media Player.... the list is endless. It is slow to load the first time off the HDD, but once it is in RAM, it is blazin fast. I don't even see the splash screens.

An analogy:
Imagine you have a lot of boxes you want to put in storage. It would be good if your storage facility had wide doors, in house moving equipment, space efficient. This is what the benchmarks test. They are all dealing with relatively small block size and how well the RAM operates.

The only way to store more boxes is to get more storage space. More RAM will help that. The more objects you can cache the better performance you will see during that session. THAT is the truth!

Also oops, I don't think that people using this forum are your basic users who only check email. I suspect they are a little more hardcore than that.

 

[su root]

Also oops, I don't think that people using this forum are your basic users who only check email. I suspect they are a little more hardcore than that.

you prove that yourself:

... ACAD2000, UT2003, TacOps, IE, Word, Excel, Access, Outlook, Matlab, Mathematica, Windows Media Player.... the list is endless.

.. i also suspect alot of people here do video, sound or graphic editing aswell, and those programs take up gobs of memory.

To see if more ram is right for you, fire up performance monitor, and put a monitor on Memory Free, and HD Page Requests. If Memory free is often very low, or HD page Requests is often high, then you need more ram (kuz your system is running out of ram, and is paging... this can be seen by the "thrashing" of your hard drive when doing alot of stuff)

Personally, I keep alot of stuff open.. i have 384mb/sdram right now, i'm going to upgrade to 640mb/sdram soon, as I see my system thrashing during high-use periods.