Using your assumptions case, you might as well leave the pagefile on the primary drive since the secondary drive is ALSO likely to be used at some point in time.
It’s been well proven that putting the swap file on a HD other than the primary (OS) drive makes a massive difference in performance.
It won’t work in every case, and indeed, if you are thrashing your secondary HD for Video input/output, the virtual ram on that HD is next to useless.
Now consider … you have virtual ram spread across 2 HD’s, a secondary & tertiary … the secondary is being used for video input/output (and has 1/2 the swap file) and the secondary just has 1/2 the swap file.
What will give better performance? 1/2 the swap file on two HD’s, or the entire swap file on the HD you are thrashing for video?
At any rate we are no longer in the dark ages.
Sata -> Every drive on it’s on channel.
Sata controller connected directly to a fast (non-pci / non-bandwidth limitting) bus.
Currently-> Mixture of PATA & SATA is very common on motherboards.
Arranging each HD to have it’s own channel is not as difficult/expensive as it used to be.
I’ve currently got a 1x120GB IDE HD, 2x160GB Sata & 320GB Sata (which is obscene).
Again, I have to re-iterate, you only use the virtual ram when the system/porgram runs out of real/allocated ram. If you have plenty of ram (ie>1GB generally), and most recent machines do, it’s a moot point
The only time you will notice it is when you are running multiple programs and swapping between them frequently.
No. In a two drive situation it makes sense to have ‘entire’ swap file on drive D if most of the work is done on drive C: and the slaves are free. Even better is to have a drive dedicated to ‘only’ swap. Likely in our cases though (being CD freaks) Primary slave and secondary probably has opticals so there’s no perfect solution and plenty of scope to actually make things worse in a two drive situation on IDE channels.
But that’s deviating from your case in which you have ‘multiple’ shared swaps. From my count you have 5 including one on a slow USB drive
I don’t know about “massive difference” but it can be be beneficial under ‘certain’ circumstances depending on configuration and use. The argument I make is whether the difference is worth the bother and can be undertaken without going from bad to worse or creating other nuisances as a result.
Whether or not we agree on how much performance difference it makes it’s a fairly irrelevant and moot point to argue about if you maintain the system only uses page file when it’s out of real ram. In which case having swap files located everywhere is a messy and wasted allocation of drive letters to have to put up with for a very infrequent benefit when and only on the odd occasion your system runs out of ram.
Ok, from what I can gather from above quote you’re talking about a 2nd and 3rd HDD, though your use of term “secondary” has implications of talking specifically about P/ATA and IDE channels which have limits and implications of their own I’ll not get into to muddy the waters.
Anyway, regardless of where drives are located on system resources: If I’m correct in my assumption of your example and your ‘terms used’ then you actually have 3 HDDs in use. 1. being OS drive without any swap. 2. being drive with half of swap file but also containing video. 3. being (tertiary which literally means third) drive containing other half of swap allocation with unknown content and use?.
In such an example either half ‘or’ total swap file is going to be a performance penalty when placed on the second drive used for video. Of course having the entire swap file on drive 2 is the worst case scanario. However, under either situation the entire swap file would be better on original OS drive or on the third drive otherwise not in use. It make no sense whatsoever to have any part of a swap file on a drive used exclusively for video whether it be for storage, capture, periodic use, or network streaming. The sequential and prolonged I/O nature of such use is going to severely impact performance if paging needs to be done on same drive during ‘any’ access or writes of video. If you were capturing raw video to such a drive you would definately lose frames if any part of a page file was used on such drive.
Another point is the nature of a single drive that is used for video: Ideally cluster size would be large at 32k or 64k while page file is best at by nature of writes and performance at windows standard 4K cluster size on single drive. Create a seperate dedicated partition perhaps, but still the severe interuption occurs to heads otherwise engaged in prolonged sequential use as well as having to cross free space between partition boundaries to use the page file.
From your configuration it seems you have dedicated partitions for your swap files, ‘possibly’ even on drives with hidden partitions that contain other OS’s to boot from. If so, the intangible benefit of seldom used swap files (by your interpretation of use) for windows is undoubtedly causing a performance penalty for any other OS installed. I say this because there’s got to be a trade off somewhere. Either the windows swap files are located on fastest part of platter (start) in which case any other OS takes a hit in being installed with it’s own swap further into the platter, after the windows swap partition, or the windows swap partition comes after something else, which defeats purpose of putting it there being on slower part of platter. Either way it’s rather pointless to be on such a drive if the idea is to improve rather than impede performance of any OS used.
^^^ I can only guess at your use and partition arrangement.^^^
Well I did state “Such depends on what the drives are attached to, the available bandwidth, what else is attached to the same bus etc etc”
Even todays chipset architectures are less than ideal and varied in configuration and they ‘still’ have bottlenecks and compromises. Personally I hate the 16 lanes allocated arbitrarily to graphics on all todays boards. It’s a waste to anyone that doesn’t play games on mega $$$ graphics cards and even then the available bandwidth isn’t likely saturated.
If you were talking about relocating the page file in it’s entirety to a Raid 0 array or Raptor seldom in use hanging off PCIE lanes then I could see it ‘possibly’ being worth the effort. Otherwise it’s not, IMO
While I disagree with your assertion whether re-interated or not that windows will only turn to “virtual ram” when it runs out of “real/allocated ram”, it’s obvious you believe it which is what’s important. So, given that case, the question remains why bother with multiple swap files for very infrequent use or use that may not occur at all?
^^Be easier and far more organised with benefits clearly not in contention to merely increase memory and regain all those wasted drive letters. But then we wouldn’t of had this marvelous topic to discuss