100

What I understand about 32-bit OS is, the address is expressed in 32 bits, so at most the OS could use 232 = 4G memory space -- I assume the unit is bytes, so 4GB.

Does this mean if any machine with a 32-bit OS (be it Windows or Unix) has more than 4GB total of RAM + page file on hard disk, for example 8GB RAM and 20GB page file, its memory will never be "used up"?

By "used up" I mean that increasing RAM or page file won't help the performance; of course, it's always possible an application will keep requesting memory from the OS but failing.

Similarly, if this 32-bit OS machine has 2GB RAM and 2GB page file, increasing the page file size won't help the performance. Is this true?

Improve this question
3
  • 2
    As a side note, if you happen to have more then 4GB of RAM, and only a 32 bit OS, you can make a RAM Disk. This is a hard drive in your RAM, and you can use the unused/unaddressed RAM to place it in.
    – Simon Verbeke
    Dec 13, 2011 at 13:33
  • 10
    You can't. A ramdisk still has to be created throught the OS, if the OS can't use the memory you can't create a ramdisk there either. (Unless your bios can create a ramdisk, which is rare).
    – AVee
    Dec 15, 2011 at 8:17
  • 1
    The short answer is "yes". Although usually a process is given a 32 bit address space to use, the OS itself can use more than that (either by bank switching like we did in the 80s or by address extension tricks school as Intel's PAE. Also, remember you run more than one program at a time, which may be composed of more than one process (Chrome puts tabs in different processes). And the OS itself will use the memory for things like caching data from local fixed disks so that you experience faster access to data already used.
    – rbanffy
    Jul 14, 2012 at 13:20

4 Answers 4

Reset to default
148

What I understand about 32-bit OS is, the address is expressed in 32 bits, so at most the OS could use 2^32 = 4GB memory space

The most that the process can address is 4GB. You are potentially confusing memory with address space. A process can have more memory than address space. That is perfectly legal and quite common in video processing and other memory intensive applications. A process can be allocated dozens of GB of memory and swap it into and out of the address space at will. Only 2 GB can go into the user address space at a time.

If you have a four-car garage at your house, you can still own fifty cars. You just can't keep them all in your garage. You have to have auxiliary storage somewhere else to store at least 46 of them; which cars you keep in your garage and which ones you keep in the parking lot down the street is up to you.

Does this mean any 32-bit OS, be it Windows or unix, if the machine has RAM + page file on hard disk more than 4GB, for example 8GB RAM and 20GB page file, there will never be "memory used up"?

Absolutely it does not mean that. A single process could use more memory than that! Again the amount of memory a process uses is almost completely unrelated to the amount of virtual address space a process uses. Just like the number of cars you keep in your garage is completely unrelated to the number of cars you own.

Moreover, two processes can share non-private memory pages. If twenty processes all load the same DLL, the processes all share the memory pages for that code. They don't share virtual memory address space, they share memory.

My point, in case it is not clear, is that you should stop thinking of memory and address space as the same thing, because they're not the same thing at all.

if this 32-bit OS machine has 2GB RAM and 2GB page file, increasing the page file size won't help the performance. Is this true?

You have fifty cars and a four-car garage, and a 100 car parking lot down the street. You increase the size of the parking lot to 200 spots. Do any of your cars get faster as a result of you now having 150 extra parking spaces instead of 50 extra parking spaces?

Improve this answer
13
  • 27
    I find this answer to not address the question at hand and almost is deceptive at points.
    – Rig
    Dec 13, 2011 at 19:49
  • 50
    @Rig: Awesome, I am glad to hear your constructive criticism. Now's your chance to write a better answer and show us the sort of answer you think is less deceptive and more pertinent.
    – Eric Lippert
    Dec 13, 2011 at 20:40
  • 4
    @BlueRaja: I suspect that you are confusing physical memory with memory. Physical memory is just an optimization that permits faster access to memory. Just as processor caches are an optimization that permit faster access to memory. Memory is for all intents and purposes space reserved in the page file. Whether a given page in the page file is (1) mapped to a virtual memory address or (2) copied from disk to physical memory is irrelevant to the question of whether it is allocated or not.
    – Eric Lippert
    Dec 13, 2011 at 22:57
  • 4
    @BlueRaja: Raymond Chen's article on the subject might help. He shows how to allocate 4GB out of the page file; that is memory that the process owns. The process cannot address the whole thing at once of course, but it can address any page by mapping a page into an unused VM page, using it for a while, and then discarding it. blogs.msdn.com/b/oldnewthing/archive/2004/08/10/211890.aspx
    – Eric Lippert
    Dec 13, 2011 at 22:59
  • 3
    This answer would have been clearer, Eric, if you had stated explicitly that Windows (and other OSes) allow you to “reserve” memory without actually having it in your virtual address space, and to then answer the obvious question “then how do you address it?”. I think the answer to that is “there is some sort of Page ID that you use to refer to it in order to page it into the virtual address space, and then you can address it there”, but I don’t know, and your answer doesn’t say :)
    – Timwi
    Oct 24, 2012 at 15:20
69

It is true that the CPU can only address maximum 4Gb of RAM. However, current CPU's use an MMU (Memory management unit) to translate process-specific memory addresses into physical memory addresses.

This MMU is used for all sorts of different tricks, from memory isolation (process A cannot manipulate memory of process B) to memory sharing (process A can access the same physical memory region as process B and can exchange data this way).

Although 32-bit CPU's only support 4Gb of memory per process, it can address up to 64Gb of RAM when using Physical Address Extension. This allows process A to use the first 4Gb of memory, while process B uses the next 4Gb. In total, more than 4Gb of physical memory is used, but the total amount of memory a single process uses is still capped at 4Gb.

PAE is supported on Linux since kernel version 2.3.23 and on some 32-bit flavours of Windows Server, but not on 32-bit Windows XP, Vista or 7.

If your CPU does not support PAE you will be limited to 4GB of physical memory (or less depending on other factors).

Please note your operating system can still evict parts of physical memory to the disk (page file) regardless of the CPU supporting PAE. This ensures you can start multiple processes who use more than 4Gb combined. The only impact PAE has is whether you can keep the 4Gb of process B in physical memory while running process A.

Improve this answer
13
  • does it mean if it does not support, then what i said above it correct?
    – athos
    Dec 13, 2011 at 13:03
  • 11
    Windows XP does support PAE (starting with SP2, I think), but ignores RAM beyond 4GB even when the machine supports PAE; see here. It's used for hardware-based Data Execution Prevention — the NX/XD bit is available only in a PAE page table entry, not a "regular" page table entry.
    – Wyzard
    Dec 13, 2011 at 13:23
  • Chipset can also be a limiting factor of how much memory can be supported/used regardless of which OS or hack you choose.
    – Moab
    Dec 13, 2011 at 13:53
  • 5
    I'd add that the reason why consumer versions of 32bit windows don't support >4GB of ram was that a significant number of hardware drives were found not to work correctly with that much memory; and since at the time >4GB of ram was extremely rare outside of servers/ultra high end workstations that most of the companies who maintained the drivers were uninterested in spending money to update drivers that would only be used by 0.1% of their customer base.
    – Dan Is Fiddling By Firelight
    Dec 13, 2011 at 14:06
  • 7
    @DanNeely the OS does have support for the memory, it's limited due to licensing issues. While I agree that this is a third-party hack, it's only for consumer variants of Windows. Server-based Windows distributions, starting from Windows 2000, have had support for more than 4GB of memory (ref. the first link).
    – Breakthrough
    Dec 13, 2011 at 15:42
15

Speaking specifically about 32-bit Windows variants, they have had support for more than 4GB of RAM since Windows 2003 variants (and you can also get a kernel hack for Windows 7 to allow you to use all of your RAM in 32-bit). However, this comes at a cost, as you outlined in the first part of your question.

In a 32-bit operating system, the size of a pointer (memory address) is the same as the word length of the CPU, 32-bits, which allows (as you mentioned) a 2^32 = 4GB memory space. Windows also takes a "virtual memory" approach for applications, so each application has it's own memory space.

Since each pointer is only 32-bits wide, each application's pointers can only address up to 4GB of memory, even though the system can support more then 4GB of RAM. As far as I know, this is the only caveat to using more than 4GB of RAM in a 32-bit operating system. In total, you can have many applications using more than 4GB of RAM combined, but any one particular process can only allocate/access up to 4GB.

Back to your question, let's say you have a program that uses 2GB of RAM. If you have 10 instances of this program, that's 20GB. All 8GB of your RAM will be used up, as well as another 12GB of the pagefile. So yes, under 32-bit operating systems, it is more than possible to use up this memory.

if this 32-bit OS machine has 2GB RAM and 2GB page file, increasing the page file size won't help the performance. is this true?

Increasing the pagefile size will usually not increase performance (unless your RAM and pagefile are set to the absolute minimum, or set so low your computer constantly thrashes). It will, however, prevent your computer from running out of (virtual) memory. Whenever anything needs to be purged to the pagefile, you're already taking a huge performance hit (since the hard drive is orders of magnitude slower then your RAM).

Improve this answer
4
  • Of the 4GB of address space that each process gets, only 2GB is actually available for the program to use; the other 2GB is reserved for use by the kernel. The /3GB boot option can allow some programs to use more RAM.
    – Wyzard
    Dec 13, 2011 at 13:30
  • 1
    @Breakthrough: "Increasing the pagefile size will never increase performance" I don't think this is true. For example, if there is more swap, the OS can more aggressively swap out unused program data, and use the free RAM for e.g. disk caching. In some configurations, this may yield better performance. There is a nice answer on ServerFault on why a pagefile is important, which touches on these issues: serverfault.com/questions/23621/…
    – sleske
    Dec 13, 2011 at 13:47
  • 2
    Be careful of blanket statements like "Increasing the pagefile size will never increase performance", it is perfectly possible for a page file that is too small to result in more disk thrashing than having a larger page file! See answers and comments in this question.
    – Mark Booth
    Dec 13, 2011 at 14:21
  • 1
    @sleske Good point, I updated it to be more of a "usually not" type of thing. While I agree it may make a huge difference in a server, for most people, the point I was trying to get across is that they would be better off just getting more RAM (since when you're paging, performance is pretty much as slow as it can get).
    – Breakthrough
    Dec 13, 2011 at 15:11
-1

When a processor is said to be 32-bit, it means it can operate with 32-bit numbers using a single instruction. This has little to do with the width of its address bus, which on Intel architecture is 36-bit since Pentium Pro released in 1995.

The famous 4GB limitation comes from the fact that most PC software uses Flat memory model where each byte of memory can be addressed by a pointer. Since a pointer should fit in a register to be used, and registers are 32-bits wide, you're limited to 4GB.

Improve this answer
6
  • This is incorrect, as pointed out in other earlier answers above.
    – ChrisInEdmonton
    Jun 17, 2015 at 14:02
  • Could you point out one particular fact that I got wrong?
    – Dmitry Grigoryev
    Jun 18, 2015 at 8:05
  • 1
    A single process can use more than 4 GB. Specifically, previous answers point out "A process can have more memory than address space".
    – ChrisInEdmonton
    Jun 18, 2015 at 12:15
  • 2
    See for example, blog.superuser.com/2011/04/23/… Now, certainly, you can't use more than 4 GB simultaneously, but a single process can unmap and remap the memory, so can make use of more than 4 GB in a single process, there's just a bit of bookkeeping. That blog post is rather Windows-specific. I've run database servers with > 4 GB of memory allocated, on 32-bit Linux installs. These days, of course, you'd just run 64-bit Linux.
    – ChrisInEdmonton
    Jun 18, 2015 at 14:16
  • 1
    OK, my apologies. I didn't think about the possibility to mmap extra memory.
    – Dmitry Grigoryev
    Jun 18, 2015 at 14:31

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .