VOTE: where to place your virtual memory? (swap file)

It is hard to say without testing. I think personally there is to much math to consider because memory structure in the computer is pretty complex there is all sorts of math to consider.

I think an easy way to think of going about it is this.

- If the systems committed memory which is a sum of all physical ram + virtual ram(page files) combined, is never saturated then there will be very little paging. This day and age with common ram amounts being at least 32GB and higher then the committed ceiling is quite high.

- Every drive on the computer should have a page file, its a round robin effect so the OS will pick which ever one responds first meaning each drive adds redundancy.

As for performance there is way to many factors and i think as long as you have a good committed amount of memory that is not saturated then there is going to be no performance impact because the page file is not really being utilized to a level that would impact performance.

There is way to much math to it tbh and performance is not going to be on the table because it only effects performance really when the committed memory is backlogged and saturated because things are waiting to move in and out of paging, the compute becomes unstable and things might crash.

There is plenty of math to consider its pretty complex situation, Nobody said anything about moving the page file it was about having it on another drive not sure what your point is tbh. It does pick which ever page file responds first.

we are essentially talking about memory management where the page file is supposed to be a fail safe system here so Stack size, virtual address range, committed pages???? Alloc, Heap and other areas of any given application/s, The memory and its whole paradigm overall? yes its pretty complex situation and yes there is plenty of math to think about when talking about memory.

Imo post 11 stands and is logical way to think about the situation, as long as you have enough ram and the system is not capped to its commit capacity then there is not much reason to overthink it because outside of that it gets complex real fast. The above are all hold bearing when talking about memory but to someone who is not writing an application or even dealing with memory management then you need to think about it in a easier fashion.

The goal should be to not page at all which will happen if you have enough ram, the only thing going into page is system data, it stores driver information there and other system parameters outside of that no one wants their applications paging at runtime because its extremely slow.

IDE is not SATA either they are not even remotely similar in terms of functionality. The only thing similar is that SATA was the successor to PATA but they are pretty different in scope and the paradigm changed considerably between the two models.
Well on some level they are similar but its like comparing a modern graphics card to the first 3D accelerator AGP card they are to far removed.

SATA came and its tech overshadowed previous methods and it pretty different radically compared to PATA, is 16bit so 16 data channels and also running in half duplex so 8MB data transfer rate where the first iteration of SATA is 1.5Gbits data transfer rate over two channels for transmitting and receiving.

Right, wrong or.....

The answer is always.... It depends on.
I can make this a technical long essay that 60% of the forum users need to google on fancy words to grasp what i mean.. But i will keep this basic as i did when i worked as a teacher assistant.... as i like new people that dont is in IT to understand forum conversations.

So i started... The answer is always.... It depends on.... Why?
First, what kind of hardware do you have.. state of the art professional workstation, or an 15year old laptop with one 5400rpm HDD.

Second, what do you use the computer for. hard rendering videos and hi data calculation usage.. or do you only check your email and one or two youtube videos.

I am a power user and I voted............. Separated: one for OS and data, and another for SWAP
Why did i do that.. It applies well on all my systems.

I keep my OS on a separate drive and Data on a separate and also the swap on a separate.. This goes for both windows and Linux, and with laptops that has one drive.. then its Partitions instead of disks

This way it is easy to back up the OS with disk images and the images dont contain a ton of data and i dont have to customize imaging to avoid getting data on the OS image..

With data on a separate drive i can save/download data without even thinking if my drive gets full.. if it does, nothing happens and the system is still ticking on.

Swap on a separate drive, same here if i do something stupid as filling the OS drive to the last bit. The system dont freeze and i get a small window of empty the system drive before the system starts to save to system disk.
But if i have Sata disks.. HDD/SSD i also gain performance om having separate disk channels as Sata is simplex.. Like old networks where you could only send or receive data in sort burst at the time as you could not both send and receive data at the same time.
So if you working with something that you need swap as you dont have 1024GB of Ram and you working with data on disk.. You can not both read from disk and send at the same time and you get a bottle neck.

Say you have a laptop with one HDD and NTFS file system.. you also have use of having swap on a separate partition to minimize fragmentation on the disk.. Sure you can get kind of fragmentation on SSD too, but that do not matter as the disk is soooo much faster that fragmentation do not impact performance and you dont store files in the same matter as you do on an spinning disk.. but that is another topic.

Say you have a state of the art computer with 10 nmve disks and two 12 SAS SSD's pools in a Raid10 setup and 1024GB of Ram... and the only thing you do is to check your email and a few youtube videos... then with still using the principle one for OS one for Data, and another for SWAP it still works and you have not sabotaged or made your system noticeable worse.

.... Am i wrong?





Edit
It might take some time before i reply on quotes, as I'm off to bed now and i am busy tomorrow, so i might look in a few hours before bedtime tomorrow.
Take care folks.

Q: Where do you place your virtual memory?

A: I don't remember.

Malneb said:

IDE is not SATA either they are not even remotely similar...

IDE describes SATA and PATA...Parallel ATA - Wikipedia

Again, the SWAP file can't be changed to pick and chose what ever drive you want without symlinking it...

https://social.technet.microsoft.com...w8itproinstall

The use case for a page file has diminished exponentially since we have been breaking into x64 more and more over time which just basically means on a easy note surplus amounts of memory and eventually it will be adhoc use case and then it will be redundant realistically it is all of these right now depending on use case situation.

The system will store data in the page file and some apps may store small amounts there too if there is need to do that but it is a dying paradigm overall and really is a throwback.

I would think that its good line of thought to have one just to have one but it does not need to be overly big either as that image shows there is no real bearing on needing one i was running a game this browser with multiple tabs some office software, 7 various apps in my tray and there is all very little paging being done because i have a big physical memory pool.

if i was using 4GB ram or less then yeah page file becomes a necessity on a 1:1 memory to page size ratio because it would be easy to fill up that sort of ram with windows average around 2.5gb alone then that would be half of you cap right there.

We could have no page file if we want i used to do this, if i ever saturate my ram in that case which could happen but was rare then i don't mind a crash and windows will pop up a memory error so i know the computer will crash any minute unless i manually free some ram by closing apps, risk vs reward scenario and in those times you don't care because trade off.

I have one now because disk space is plentyfull but it does not need to be big either because we are not really relying on it like that in 2023. Going back 25 years and older drives and ram where still pre 1gb threshold. That is no longer the case.

Everything in the computer is also going on the CPU because it needs to go that way for progress so with CPU cache and FSB being on the CPU then there is little wait time and things don't hang around waiting that long they go in and out or Ram, CPU hard drive pretty fast now.

F22 Simpilot said:

IDE describes SATA and PATA...Parallel ATA - Wikipedia

Again, the SWAP file can't be changed to pick and chose what ever drive you want without symlinking it...

https://social.technet.microsoft.com...w8itproinstall

Yeah but no one even mentioned about moving the swap file its clear the thread realistically is about page file because you can choose where that one is located i would say that there is confusion here. The two terms used to be synonymous and interchangeable because they were the same thing initially and still are just different names over different systems.

swapfile.sys is for UWP apps where they can write the whole memory state to that file it was made for this very reason and is still the same sort of use case as the page file, although I think its irrelevant and out of context to undermine here because the OP is confusing the two but we know what they are meaning.

for the sake of the thread just use the correct term page file because that is what we are really talking about here and that is what its always been called in windows. Swap file is for specific programs in windows and is relatively new where its not really what we are talking about at least i don't think so.

You also came in here looking for an argument i think, i should not have to defend myself i was not trying to talk over or disprove anyone just sharing my thoughts.

Yes SATA comes from the IBM AT Standard line of evolution if you go backwards but its not really relevant to state that they are the same thing along the way because they are not.

PATA used controllers on the hard drive where Sata is using AHCI and now NVME tech which are very different. The whole paradigm has changed. NVME especially is now like uncharted land compared to PATA

I had more need for manipulating the Virtual Memory [VM] in years/versions of the past but computers have gotten so fast I go with the defaults. Older drives being slower worked better if files were not fragmented to which end I'd turn VM off, defrag the drive, re-enable VM at a fixed max and min size which kept it from getting fragmented, had noticeable performance improvement. Also, a number of programs during install will look for a VM on the same partition as the OS, didn't have to be large, maybe about 2GB. I never put the VM on a second partition the OS is on it has to use the same Read/Write heads of an HDD to work with both partitions, could slow things down. An SSD is a different sitiation.