I found that article pretty disappointing. "Where's the beef ?" The article talks about a test setup but they did no real tests. Sure they talked about what happens if ya turn the swap file off and "yeay we had no problems at 8 GB". But again, where's the beef ?
Where's the benchmarks ? Where's the results with 2 GB, 4 GB, 8 GB ? I don't really wanna know that "yeah well theoretically of you don't have to write to the HD, it should be faster". I wanna know the numbers. They said that Windows ran fine at 8 GB with no swap file but at 2 GB they ran outta of memory loading 3 GB files....well duh !
They also say "Our testing brought us to a clear conclusion: if you often use several memory hungry applications simultaneously, then there's really no way around upgrading your system to 8 GB. Working with applications, and especially switching between them, is much more efficient than with a typical 2 GB configuration."
Again, 1) I saw no test results and 2) "no way around upgrading to 8 GB ? I see nothing explaining why 4 GB is this huge roadblock.
All I got out of this article is "I bought 8 gigs of RAM and here's my attempt to justify it". If it's better it should benchmark better. And if it's multitasking argument, I wanna see what the tester is multitasking. And not some silly "I downloaded SP1 while running a backup, while running writing a DVD while running my AV scans while defragging my drive, while running a find and replace in a 2 GB word file wile editing a PhotoShop Image while compressing a movie...."
Machines can multitask but humans can't. While it is valuable to be able to run programs in background while getting work done, the test should have some semblance of reality. An intelligent user, runs AV scans and backups at night for example. And if you are editing a movie while editing a photo while editing a word file, you are not doing any of those things well.
I have no problems with normal things.....like having some programs open and playing a game....that is something that happens everywhere....I routinely do that during lunch at the office. I fan PCMark and Futuremark with a many as 10 programs open including Adobe Acrobat and Designer open, AutoCAD open others and I still failed to generate significant differences in benchmark results between having all 10 programs running and having all 10 programs off.
What the reviewer failed to acknowledge is that these page file writes are not repeating which is why I figure he didn't back up the claims with benchmarks. Let's say you have those 10 programs loaded .... even when you load the first one, windows will stick some of it in Virtual memory (VM). Right now I have over 1 GB of physical memory free. Yet:
AutoCAD 360MB in memory / 356 MB in VM
Firefox 119MB in memory, 95 MB in VM
The assumption that Windows will not offload memory to VM if there is physical memory available is a myth. When I load 3D Mark, the system may offload parts of those 10 running programs but then those programs are simply idle. There's no swapping going on for the next 15 minutes or so while the benchmark completes. The only way multitasking generally becomes an issue is if the machine is being forced to switch its focus from task to task 2 task 3 every few seconds and the reality is very very few users find themselves in that situation.
I have always over bought RAM. I never put less than 2 GB in an XP box.....never put less than 1 GB in a Win2k box, never put less than 512 MB in an NT4 box and I'd never put less than 4 GB in a Vista box. But as the OP said, he's never seen his system use more than 2 GB.....to which I'd say "then you should have 4". Cause if you just had 2 GB, you'd see Windows paging more cause it always seems to wanna have half the RAM free. It's pretty simple to see if 8 GB will benefit you.
-With 4 GB in, open all the programs you are likely to have open on any given day at one time. Run PCMark and 3DMark and anything else you wanna run. Record the numbers.
-With 8 GB in, open all the programs you are likely to have open on any given day at one time. Run PCMark and 3DMark and anything else you wanna run. Record the numbers.
If there's not a statistically significant difference between what you recorded under both scenarios, then the 8 GB isn't helping ya any. For a mainstream Vista PC, I'd be more inclined to invest in memory with faster timings at 4 GB then boosting run of the mill memory to 8 GB. There are instances where the extra 4 GB will help. For example, when multiple virtual machines are used, such as those provided by VMware and Microsoft Virtual PC, we are outside "mainstream user".
Finally, the author's comments on "issues" with using 8 GB should be considered:
"Nonetheless, even veteran users should expect to encounter a few setbacks when using 8 GB of RAM. Be prepared to encounter driver problems, and not just on exotic hardware. Some system tools may also spring a nasty surprise on you."