You’re confusing memory with RAM. A program or process allocates memory, not RAM, in its own virtual address space. In general, it has no idea whether that memory currently resides on disk (paged out) or in RAM. It’s not the process which controls RAM allocation, Windows virtual memory management decides what pages should be in physical (RAM) page-frames.

Leaking (a misnomer) is when a program or process loses track of memory (not RAM) which it has allocated. It isn’t a “necessary evil” at all, it’s a programming error. That memory is allocated in the process’s own address space, no other program or process can fix that situation, and the memory is freed when the process terminates, because XP and higher will free that process’s entire address space. As to when Windows gets around to reporting the freeing of RAM associated with that process is up to Windows, but it knows the RAM is no longer in use and will use it for other processes as needed. “Leaking” is generally pretty harmless. The process’s allocated memory will continue to grow, but that’s in its virtual address space, so it will wind up on disk if RAM is needed for other processes. And as I said, it will all be deallocated when the process terminates.

How Windows handles virtual memory has nothing to do with the age of the application or how it’s coded.

As for the particular Mark Russinovich article someone linked, it only costs $6 for a monthly pass. As for their privacy policy, it’s the usual “We respect and value your privacy. We only give your personal information to our carefully selected partners, which are anybody and everybody who pays us for your personal information.”

Mark Russinovich is a very smart guy, but comparing him to Einstein? Russinovich has spent his career reverse-engineering Windows, since nobody at Microsoft knows anything about Windows. He provides all the tools that Microsoft should have written. So they hired him. He’s a solid nuts-and-bolts programmer who’s based his entire career around Microsoft Windows. Is he one of the programmers who created a new programming language, operating system, or something creative like webmail or social networking software? No. He’s a good programmer, but he isn’t great. Einstein merely a visionary who screwed stuff up? Scientists generally rank Newton #1 and Einstein #2 among the greatest scientists in history, and the more I learn about Einstein, I think that he ties for #1. Physicists are limited by the knowledge and technology available during their lives. Both Newton and Einstein made some mistakes, as everyone does. Einstein definitely wasn’t “wrong in many of the particulars”. Despite some claims to the contrary, he was a great mathematician. Experiments during his lifetime and since have matched his mathematical predictions, completely supporting his theories, upon which all of our modern technology is based. It’s been said that at any time there are only around a hundred people in the entire world who can fully understand Einstein’s math (and I don’t mean plugging values into the end-result simple equations). Einstein’s ability to explain things in a way that the rest of us can understand is an amazing part of his genius, but only a part.

Apart from the links that #132, da, provided, there are Mark Russinovich’s technical blog posts on
Physical Memory
Virtual Memory
Paged and Non-Paged Pool
in addition to various Microsoft articles, which I’ve linked in the past.

More detail on what I posted earlier about Vista pagefiles (also applies to XP): Use fixed-size pagefiles. Telling Vista to manage the pagefiles itself causes it to delete them upon shutdown.
#160, Spiritwolf, increase the size of your pagefile(s).
Turning off paging is a very bad idea, and will generally adversely impact performance, even if you have a lot of RAM.

That’s the whole point of leaking. A phenomenon seen less and less, but that even Russinovich acknowledges in at least one of the links that have been posted here in the past. Leaking is a necessary evil caused by the need to make intelligent decisions about what code/functions/modules – regardless of term used, you get the idea – need to be close at hand. Windows has become better and better at managing this, especially with the implementation of virtual memory. However, older software and some software that isn’t coded in a way to take advantage of Windows’ recent innovations can still leak. Though, again, this is an ever decreasing problem.

So, whether you get RAM back for any length of time, depends on what was purged to the page file in the first place. And those who say it will always slow you down by constantly thrashing the hard disk, again, don’t look deep enough. Again, it all depends on what’s been purged. As hard as it is for some to grasp, Windows will NOT waste time pulling code back out of the pagefile that you’re not using anymore, and won’t be using for the rest of the session. That’s why is so difficult to predict whether this will work for you at any particular point in time.

Also, the Russinovich link that everyone posts is to a fragment of an article that requires a subscription to read. Does anyone actually have such a subscription? Has anyone actually read the whole article to check, an hopefully learn from, his logic? Don’t be too swayed by authority folks! Russinovich is execeptional. I’ve been referring to him for years, even before I realized HOW exceptional he was. As a former journalist, among other things, I can tell you that you should read every article to try and evaluate whether it’s logic makes sense. Does it’s bucket hold water? For example, many physicists will admit that Einstein was actually WRONG in many of the particulars. His brilliance lay in his INCREDIBLE vision.

Also, to clarify. The claim that this will speed up your system is indeed bunk. I’ve never said that, myself, even when I used these softwares in the past. The BEST it can too is help hold off the slowdown that results from being very low on RAM. But, as I said in my first post, there are better ways to do that.

As for your particular performance issues, see my comments #123 and #125. By default, Vista “deletes” the pagefile, in Microsoft’s words, “for security”. This is bunk. Microsoft doesn’t overwrite the pagefile, which would provide security, and eliminate the need to delete it. Why does this matter? Because the pagefile can’t be defragmented while Windows is running. My impression is that Vista re-uses the “deleted” pagefile, which can’t be defragmented in the deleted state, therefor it remains fragmented. Set Vista to keep the pagefile, and use a defragmenter which defragments pagefiles. Vista itself has a defragmenter, but it doesn’t do pagefiles. There are some freeware defragmenters, but the only defragmenter I trust is the commercial Raxco PerfectDisk, which is often available at a discount (I haven’t posted them lately). Their code quality has gone downhill, I’m going to have to get on their case, but that’s related to scheduling and other issues, it won’t screw up your PC. You can download a trial. Set your security software to scan on execute, not file access (at least during the defragmentation). Do a full defragmentation of your drives first, then do a boot-time defragmentation of your system files (after telling Vista to keep the pagefile). I’ll bet your performance issues evaporate. Some things can still cause brief pauses; I sometimes experience a brief pause when my security software updates.

#166, CharlesKane, that’s an excellent article for the many GOTD users who clearly don’t know what they’re doing and commented on this giveaway. I disagree somewhat with some of the article’s comments on disabling services. SuperFetch is useful. Windows Defender is useful and doesn’t interfere with other security software, so I’d only disable it if you’re really low on resources. Windows Search (especially 4.0) is great, but if you’re using another search solution, it would make considerable sense to disable it. However, it’s unlikely that you’d notice any performance increase. It operates at low priority during idle time.