Is more RAM always better?

[mechBgon]

Big-picture time 🙂 The OP has a pretty good grasp on the situation and should consider getting a third matching 512MB PC3200 module while the prices are good, if he anticipates the need for it (heavy image-editing, EverquestII, maybe some types of video/audio editing, gaming while encoding, etc). I've run 1.5GB in three different nF2 boards and it was nice to have, since I had actual uses for that much RAM.

I wouldn't run a 166MHz-bus CPU with the memory out-of-sync at 133MHz to get just another 256MB of capacity, because the lack of synchronous operation hurts. I did a UT2003 Demo benchmark run with my 2500+ @ stock 166 bus, and the RAM at both 166 and 200. Since my video card was a mere Radeon 7500, I ran the test at 640 x 480. The Botmatch fps score for the out-of-sync setup was about 15% slower and even the Flyby score came out over 10% slower. So there's one benchie for Blain 🙂

pspada did a set of UT2003 Demo benchmarks with his nF2 Ultra400 board, to determine whether it was important to balance the memory loadout between the two channels. To quote his result:

Ok, with the 512 stick in the offset slot (Bank 0), and the 2 256 sticks in the 2 close together slots (Bank 1), so the memory is "balanced", UT2003 Citadel botmatch is:

9.461353 / 31.059847 / 84.692574 fps -- Score = 31.087503 rand[23108]

I then swapped the 512 stick into the middle slot, putting that 256 stick into the offset slot, and the UT2003 Citadel botmatch is:

9.418547 / 31.061878 / 88.517540 fps -- Score = 31.089642 rand[23108]

so, I got a better score with the memory "unbalanced".

So that was enough for me to never worry again about balancing the memory from channel to channel, no matter what might be happening underneath 😉 and also to conclude that nForce2 is my friend, working nearly equally well with any memory configuration. Just avoid running the memory out-of-sync with the CPU bus, even if it means running PC3200 at PC2700 speeds, and you're on the right track 🙂

 

[JBT]

If you need to ask a 1 GB is PC3200 is going to perform better than 1.5 GB of RAM running at PC2100.

 

[Cygni]

ive had similar results to Bgon when playing around with FSB and ram settings. Something not many people have played with, however, is running an FSB far higher than their ram speed, something i tried out pretty completly with some sticks of already owned PC2100 and an unlocked Applebred Duron.

Unfortunatly, i dont have the raw results with me on this box, and wont have access to them again until January. I believe that the ending result was that at equal Proc speeds, better performance was offered from running a 200fsb with a 133 ram clock, then by running a sync 133/133 clock.

Also something not many people have really talked about is the very limited advantage that RAM speed seems to actually have on performance with the AXP/Socket 7 platform. In my experience, even the move from SDRAM->DDR had limited advantages. With a K7S5A and a 1Ghz Athlon way back when, the move from 256 of PC133 to 256 of PC2100 caused only a 15fps bump in Quake 3, from 180 to 195 (at 640 to stress cpu)... an 8% gain at fairly high cost at the time. I will be picking up some PC3200 this Christmas, and will make a comparison of 200/133 vs 200/200 clocks on the Applebred system, and i think the actual gains will be limited, but we shall see. 😉 Even so, I very highly doubt the addition of an extra quarter gig of RAM will be worth the hit from PC 2700 to PC 2100, when you get the 166 CPU.

 

[Philippine Mango]

I don't think he cares anymore/will take our advice.

 

[Blain]

Originally posted by: Blain
>> Benchmarks <<
I want to see benchmarks for different memory configurations under XP and/or 2000! Show Me! :laugh:

Come on... Somebody help the kid out with some links to these benchmarks! :laugh:

 

[mechBgon]

Originally posted by: Blain

Originally posted by: Blain
>> Benchmarks <<
I want to see benchmarks for different memory configurations under XP and/or 2000! Show Me! :laugh:

Come on... Somebody help the kid out with some links to these benchmarks! :laugh:

Four posts up, those ones any help? 🙂

 

[Blain]

Sorry, but I'm looking for benchmarks running various amounts of memory.
I do see that I should have stated... "I want to see benchmarks for different memory configurations under XP and/or 2000!" clearer. 😱
BUT in my defence, the title of the thread IS... "Is more RAM always better?"

I remember reading a thread that had links to some benchmarks for XP, 2000 and even 98SE, using various amounts of memory. The only problem is, as hard as I search, I can't find it again.
I thought that asking about links to benchmarks might jog someone's memory. :laugh:

 

[Googer]

Originally posted by: Blain
Sorry, but I'm looking for benchmarks running various amounts of memory.
I do see that I should have stated... "I want to see benchmarks for different memory configurations under XP and/or 2000!" clearer. 😱
BUT in my defence, the title of the thread IS... "Is more RAM always better?"

I remember reading a thread that had links to some benchmarks for XP, 2000 and even 98SE, using various amounts of memory. The only problem is, as hard as I search, I can't find it again.
I thought that asking about links to benchmarks might jog someone's memory. :laugh:

I have never really had any serious problems with it.
What exactly are you trying to discover (prove/disprove) by using various ammounts of memory? Its simple add more and it hurts nothing.

 

[Blain]

Originally posted by: Googer
I have never really had any serious problems with it.
What exactly are you trying to discover (prove/disprove) by using various ammounts of memory? Its simple add more and it hurts nothing.

Actually the benchmarks I remember seeing showed a performance hit with memory past a certain amount. I can't remember if it was XP or 2000.
If I remember right, it they were benchmarks for 512mb, 1GB, 1.5GB &amp; 2GB... tested under XP and 2000, maybe even 98SE.

It was quite a while back.