perhaps there are several ways of looking at ram size / ram "speed" & cpu/ram utilization (realworld productivity)

bacillus, I respectfully disagree ... in interest of brevity, virtually all synthetic tests using identical desktop and/notebook pc's, one pr (desk+note) fitted 1.0gb ram (2x512mb) and one pr fitted 2.0gb (2x1gb) of recognized quality memory modules (standard jedec timings) will show increasing productivity (over ram size of 1x512mb) @ 1.0gb level - @ 2.0gb level only marginal productivity gain occurs (over 1.0gb) ... please note ddr1 is fitted, as ddr2 still has some teething problems (it is only so slightly faster than ddr1, it is becoming abit ackward to explain) - sdram will simply not scale (bandwidth) in same manner, so no point testing, however results similar

yes an optimized mobo chipset memory controller can make difference, but we're talking server mobo's and no note chipsets

as internal dram cell density increases, max thermal becomes critical & can only be controlled by limiting power consumption; combined w/focus on optimized dram address bus (this is fast/tight ram timings) directly cell cluster schema & dram architecture/layout ... there are really only a small number of viable combinations of these variables, w/o shifting voltage level up measureably (something avoided in all jedec std ram modules)

optimized address bus (within each chip) is simply "faster" (cooler) in low to mid density dram chips as power level does not have to be limited - at highest density level (as used on 1.0gb modules) power levels are held down, to limit max temp rise, w/resulting increased signal time in-out, even though physical paths are shorter... and now 2.0gb modules having more severe max thermal issues...

I have intentionally left out fsb considerations (& cpu cache size)

to see any benefit, photoshop (not cs) must be heavily exercised by a user determined to use all features ... shop loves 2.0gb & runs measureably/perceptably better...

for all other apps only small improvement

and have not touched on bga vs tsop(2) merits

lastly, servers routinely run 4gb or more & no one complains about slow ram - it's all ecc & hdd banks refresh constantly so ram "mass" counts for more than ram "speed"

major vid stream processing genuinely needs ram "mass", but productivity does not double at 2.0gb

edit synthetic test run individually/serially - if simultaneously multiple apps / stress tests run no improvement 2.0 over 1.0 and some tests will time out longer...

just my own view on this ...

 

Message Edited by kwokPC on 08-13-2005 11:16 AM

Yes, some people say it does slow down your system, but I don't really agree with it. Of course, there are situations where it could slow your system down, but, in my opinion, there will be no slowdowns if everything is upgraded correctly.