PAE and PSE modes can be used by 32 bit OS's that , in effect, allows them to function as 36 bit OS's.

You cannot get away from the maths of true 32 bit and true 64 bit OS's.

I think you'll find that most people refer to Windows Server 2000 Advanced as a 32-bit OS, not a 36-bit one. But, that aside, those old enough to remember, for example, the ZX Spectrum 128 will know that there are yet other ways of accessing more memory than the size of the address bus would seem to allow. The maths doesn't forbid this - it just makes it a little more tricky.

There's more than one way to skin a cat.

There's more than one way to skin a cat.

Very true; it's just me being a tad pedantic.

Hi camieabz,

I did run maxxpi though can't really understand the figures using different combination of ram. Which give better performance since the figures are almost the same.

Here are some of the results

4GB Ram
1M = (412.3 k/sec -average) (-0.5%/+0.4% - deviation from average)
2M = (352.3 k/sec -average) (-0.1%/+0.1% - deviation from average)
4M = (298.5 k/sec -average) (-0.1%/+0.0% - deviation from average)
16M = (213.7 k/sec -average) (-0.0%/+0.0% - deviation from average)
32M = (181.5 k/sec -average) (-0.0%/+0.0% - deviation from average)


3GB Ram (3 x 1GB)
1M = (412.6 k/sec -average) (-0.2%/+0.4% - deviation from average)
2M = (352.5 k/sec -average) (-0.2%/+0.2% - deviation from average)
4M = (298.0 k/sec -average) (-0.2%/+0.1% - deviation from average)
16M = (212.7 k/sec -average) (-0.3%/+0.2% - deviation from average)
32M = (181.3 k/sec -average) (-0.1%/+0.1% - deviation from average)


3GB Ram (2 x 512MB & 1 x 1GB)
1M = (413.2 k/sec -average) (-0.2%/+0.1% - deviation from average)
2M = (352.4 k/sec -average) (-0.1%/+0.1% - deviation from average)
4M = (298.2 k/sec -average) (-0.2%/+0.1% - deviation from average)
16M = (213.8 k/sec -average) (-0.0%/+0.0% - deviation from average)
32M = (181.0 k/sec -average) (-0.1%/+0.3% - deviation from average)

Thanks

On that basis there's no difference. My concern is that the test I gave might not max out the memory on testing. Not sure which tests do. Something like Sandra lite might give a better idea:

http://www.benchmarkhq.ru/english.html?/sandradl_e.html (scroll to bottom)

However, based on the results, there should be no diff, and it's probably down to how much memory you need for your setup. 3GB ought to be fine imo. Like I said earlier, you can usually get a lot more from your system doing safe tweaks (not overclocking).

Here's my laptop's status before and after tweaking:

Before

After

32% reduction in memory usage at boot time (931 MB down to 634MB). A 32.2% reduction in processes (from 59 to 40).