StarDestroyer.Net BBS

Usually what I do is set my page file to use a separate, dedicated partition at the beginning of another hard disk, since your disk's performance drops as it reads/writes data toward its last sectors. I do this on my desktop just to be safe (it is set at a custom size of 1024-4096MB, which extends to my RAM's size), since the great majority of my memory-intensive work is done on that system. On my laptop, however, I do not have a page file at all, but that is because I use an SSD for it. I hardly do anything memory-intensive on my laptop anyway, so I didn't see any real need for one.

I also format my swap partition as FAT32, since NTFS writes to the disk more often, and supposedly FAT32 is better for small volumes. Unfortunately, one side effect of this is that your page file cannot exceed 4GB, since that is the maximum limit for any file on a FAT32 volume. I haven't tried using exFAT for one, though, so that may be another option you could consider if you need that much swap space, but I don't think you will need more than that anyway.

What performance improvement, if any, do you see for all that work?

I don't have any hard numbers on hand, but I know that as data progresses from a hard drive's beginning sectors (its outer regions) toward its end sectors (its inner regions), access time for that data would gradually increase as a result from his disk's heads traveling longer distances to read (or write to) it. Applying that knowledge, if Edward was to create a separate area to host his page file, then it would make sense to host that area at the beginning of his drive.

As far as his choice of file system, I only suggested FAT32 as another way for him to maximize his swap file's access time. In reality, that part is most likely negligible for systems made within the past several years.

Dragon Angel wrote:Usually what I do is set my page file to use a separate, dedicated partition at the beginning of another hard disk, since your disk's performance drops as it reads/writes data toward its last sectors. I do this on my desktop just to be safe (it is set at a custom size of 1024-4096MB, which extends to my RAM's size), since the great majority of my memory-intensive work is done on that system.

Why don't you just let Windows do its thing normally? Sure, put pagefile space on a secondary disk if you must (while allocating enough to write a minidump file on your system disk) but artificially limiting the pagefile size is silly. And the performance of spinning disk IO is so low anyways (compared to RAM) that optimizations like this are absurd.

On my laptop, however, I do not have a page file at all, but that is because I use an SSD for it. I hardly do anything memory-intensive on my laptop anyway, so I didn't see any real need for one.

Windows' memory management model (indeed, all modern desktop or server OS memory models) require a pagefile (or equivalent) for proper functionality. Enable the pagefile and let Windows do its thing.

I also format my swap partition as FAT32, since NTFS writes to the disk more often, and supposedly FAT32 is better for small volumes. Unfortunately, one side effect of this is that your page file cannot exceed 4GB, since that is the maximum limit for any file on a FAT32 volume. I haven't tried using exFAT for one, though, so that may be another option you could consider if you need that much swap space, but I don't think you will need more than that anyway.

There is no reason to use FAT32 to host your pagefile. None.

Dragon Angel wrote:I don't have any hard numbers on hand

Then why are you giving performance advice?

phongn wrote:Why don't you just let Windows do its thing normally? Sure, put pagefile space on a secondary disk if you must (while allocating enough to write a minidump file on your system disk) but artificially limiting the pagefile size is silly. And the performance of spinning disk IO is so low anyways (compared to RAM) that optimizations like this are absurd.

Look, Edward Yee wanted some advice as to where he could place his page file, and I saw him creating a partition at the beginning of his other hard disk (non-SSD) as the best location that he could put it in. My 1024-4096MB figure that I mentioned was my own personal fiat - he could set it to a System Managed size if he wanted to with no problem, but my central point in that line was the fact that him placing his swap on a partition at the beginning of his drive would guarantee it being located at its fastest area.

Also, as a side note, I have never noticed my page file ever approaching the maximum size that I placed on it. That is not to say that it is not used at all, but rather Windows probably never actually saw a need to fill it to its limit.

phongn wrote:Windows' memory management model (indeed, all modern desktop or server OS memory models) require a pagefile (or equivalent) for proper functionality. Enable the pagefile and let Windows do its thing.

I have never noticed any problems without it, nor have I suffered any other inconveniences from its absence. Having it enabled would only cause more writes to my SSD than I would care for, regardless of how long its estimated lifespan is supposed to be.

phongn wrote:There is no reason to use FAT32 to host your pagefile. None.

*shrugs* I used FAT32 to minimize the amount of writes (journal, last access, ...) on the partition that I hosted it on. If it helps to take away some writes that Windows decides to carry out, then I view it as a small boost in speed, no matter how negligible. But as I said before, whatever file system he chooses is perfectly fine for systems in today's age. All of this is a matter of opinion.

phongn wrote:Then why are you giving performance advice?

Fine, do you want a graph to satisfy your standards? Here you go, look on the third page. It's a document describing an SSD, but that graph should elucidate you about performance degradation as you reach your hard disk's inner tracks (it was given as a comparison to SSD seek times). Oh look, here is a website with two data tables describing a performance loss with two drives. While those tables describe drives that may be approximately 6+ years old, the principle remains the same: Speed degrades as a hard drive's heads attempts to read deeper into its tracks.

I already mentioned an explanation as to why there would be a performance decrease if his page file was located closer to the center of his disk. Is a little bit of logic that hard for you to do? Would you care to nitpick my post just a little bit more?

Except Edward Yee said he'd be using Windows 7. You can't install Windows 7 on a Fat32 partition because the OS won't let you, so this entire tangent has been an exercise in absurdity.

General Zod wrote:Except Edward Yee said he'd be using Windows 7. You can't install Windows 7 on a Fat32 partition because the OS won't let you, so this entire tangent has been an exercise in absurdity.

Uh, who said anything about installing Windows on a FAT32 partition? Have you been reading what I had typed?

Dragon Angel wrote:

General Zod wrote:Except Edward Yee said he'd be using Windows 7. You can't install Windows 7 on a Fat32 partition because the OS won't let you, so this entire tangent has been an exercise in absurdity.

Uh, who said anything about installing Windows on a FAT32 partition? Have you been reading what I had typed?

Since the context of the op made it seem pretty clear he was only talking about write-minimizing his C:\ drive I can't imagine what else you could have been talking about. Since SSDs are pretty much the only thing you have to worry about as far as write-minimizing.

General Zod wrote:Since the context of the op made it seem pretty clear he was only talking about write-minimizing his C:\ drive I can't imagine what else you could have been talking about. Since SSDs are pretty much the only thing you have to worry about as far as write-minimizing.

Uh huh. Tell me, are you intentionally trying to be obtuse? Or are you just being outright dishonest?

Yee was attempting to minimize writes onto his SSD (or in that case, his C: drive), and one of the options that he had listed included him setting a page file on one of his other drives, in order to minimize writes onto his SSD! Where you somehow got "install Windows onto your swap partition!" (which, by your words of installing Windows onto a FAT32 partition, is what I can see you thought) is something that I can't understand. I guess that you really aren't reading this topic.

Rereading the OP, I noticed something that went unaddressed:

Edward Yee wrote:In this context, is it safe to treat a partition as interchangable with a separate physical hard drive? (i.e. on my three-partition HDD, set the C:\ partition for "no paging file," and the D:\ or E:\ partition for "system managed size.")

No not exactly - a partition is what it is supposed to be, just a section of the same disk that it is on. As I stated earlier, a partition at the beginning of your drive will have much faster access times than a partition at the end of your drive. If you partitioned your main disk (which I will assume to be a normal HDD here, for the sake of argument) into two separate areas, one big area for your system (C:) and another smaller area for your page file (D:), and placed your D: partition at the end of your HDD, it would suffer a definite performance decrease than if you allowed your page file to be written anywhere onto your system (C:) partition.

In terms of an SSD, laying your partitions at the beginning or end of your SSD will not matter one bit, as there are no mechanical parts in use (i.e. no latency from the distance that your drive's heads must travel). But placing your page file anywhere on your SSD would defeat the intended goal of minimizing writes onto it, so that probably isn't the path that you would want to take.

phongn wrote:Why don't you just let Windows do its thing normally? Sure, put pagefile space on a secondary disk if you must (while allocating enough to write a minidump file on your system disk) but artificially limiting the pagefile size is silly.

I remember hearing about that back when Slot 1 was still in use. Supposedly it was intended to minimize the time Windows spent calculating what size the page file needed to be and resizing the page file, but I have no idea how valid that was.

Uraniun235 wrote:I remember hearing about that back when Slot 1 was still in use. Supposedly it was intended to minimize the time Windows spent calculating what size the page file needed to be and resizing the page file, but I have no idea how valid that was.

Never even heard that mentioned. It was basically to prevent the pagefile from fragmenting, because there was no official way to defragment it. Not because the pagefile itself would be the problem, but it increased the rate at which other files on the disk got fragmented.

Not applicable with SSDs, regardless.