A Primer on Windows' Memory Management

In 1986, Intel released the 386 processor, which offered support for a new instruction set (IA32), that was an extension of the original x86 instruction set. Among the most notable features in IA32 was an increase in the amount of memory the CPU could address, moving from 20bit addressing(1MB) to 32bit addressing(4GB)(ed: this is not including the convoluted mess that was segmented addressing). All x86 CPUs released since then have supported the same instruction set, including the same 4GB limit. Only recently have 64bit x86 CPUs been released, and while they support more than 4GB of memory, still are limited to 4GB when operating in 32bit mode.

As we have mentioned in previous articles, most modern system running in 32bit x86 mode have trouble seeing and using more than roughly 3GB of memory. This is because part of the total 4GB of memory space (not the physical memory) is reserved for various functions, such as computer components transferring data between each other using memory-mapped input-output(MMIO). The textbook example of this is the CPU transferring data to the memory of a video card, where a chunk of the address space equal to the size of the memory of the video card is reserved by the video card, and any data sent to those addresses actually ends up going to the video card. This design has many technical merits, but it makes the consumed memory addresses unavailable for use with physical memory.

Things only get more complex as we start including the operating system (in this case Windows) in to the equation. The above is actually handled by a combination of Windows and the BIOS, meanwhile Windows also needs some address space so that programs can communicate with the Windows kernel, for storing buffers, for storing memory tables, etc; all of which means we have lost even more address space. All of the above besides preventing us from addressing 4GB of physical memory are also the cause of the actual 2GB barrier that is the problem.

Quickly, there is one more pre-requisite piece of information: virtual address space. For a 32bit Windows application(Win32), each application has a full 4GB worth of private addressing space that it can use, which again is 32bits and a result of how a 32bit processor works. How the above exactly works is beyond the scope of this article, but it's sufficient to say that at some point virtual addresses get translated in to other addresses for mapping data between the application and physical memory or the swap file. The important thing to take from this is that each application has its own 4GB virtual address space, regardless of the hardware the computer contains or what applications are running. Now we may begin to understand the 2GB barrier.

Due to design reasons outside the scope of this article, Windows takes a pre-determined portion of each application's virtual address space and reserves it for itself. Windows uses its portion of the virtual address range for all of the address needs listed above. At the end of the day, and what really matters, is that in designing Windows Microsoft opted to split up the virtual address space of an application in half; 2GB goes to Windows (kernel space) and 2GB goes to the application (user space). Under normal circumstances this 2GB of space is all a 32bit application has to work with, this is the 2GB barrier and as we'll see is the cause of the problems with Supreme Commander.

It is also worth noting at this point that virtual address space is not directly correlated to physical memory size. Windows and any applications running under it may use up to all of their allocated virtual address space, with Windows simply swapping data between the hard drive and physical memory if the amount of physical memory needed is in excess of what's available - this is virtual memory, not to be confused with virtual address space. The only meaningful relation between virtual address space and physical memory is that the amount of physical memory an application can use can never be greater than its virtual address space. The user space must take up a larger portion of the virtual address space if an application is to use more than 2GB of physical memory.

Index Removing the 2GB Barrier
POST A COMMENT

69 Comments

View All Comments

  • titan7 - Thursday, July 12, 2007 - link

    There is nothing you can do. If you ship a game you could test it and ensure it doesn't run out of memory (e.g. GameCube has just 24megs of system memory and the last Zelda looked great. Quite a ways away from 2048megabytes PC games run into!), but what about a mod?

    When the application starts just allocate 512 megabytes or whatever you feel is reasonable. When new throws an exception free that memory, display a warning you're low on memory and need to upgrade to Vista64, and continue. When it new fails again one microsecond later you're screwed so display a message to the end user along the lines of "I told you so!" End users really like that type of thing. ;)

    You could get a bit fancier by replacing all units with simple cubes or something, but all that does is delay in the inevitable a bit longer.
    Reply
  • ncage - Thursday, July 12, 2007 - link

    1) First step is to detect which OS you are using at startup. Is it 64 bit os or not?

    2) You SHOULD never code your application around a 2GB memory limit. It is very bad coding practice. Going through thorough examples in a short post like this isn't very practicle

    3) Some higher level langauge/constructs abstract this away from you. For example if you are using .Net CLR you don't really have to worry about this unless maybe your doing some crazy pinvoke stuff which in most cases you shouldn't be doing anyways. Of course if your doing VB6 your never going to get around it anyways because vb6 is only 32 bit.

    4) If you are doing C++ or assembly with windows then you can use the GlobalMemoryStatus() Function to Effectively see how much available address space you have.
    Reply
  • Ryan Smith - Thursday, July 12, 2007 - link

    The key to any of this is monitoring how much of the virtual address pool is in use; there should be an API call to ask Windows this. The easiest thing to do would be to give a warning at 1.9GB or so and then either do nothing, trigger a crash early, or attempt to reduce detail or flush space to stay below the 2GB barrier. The warning is the easy part, the hard part is preventing the crash, and I don't honestly believe anyone can or will be preventing crashing. Reply
  • yacoub - Thursday, July 12, 2007 - link

    I just wish we had a better solution than Vista. Sure we can use 64bit XP but that's only going to last how much longer with full patch support from MS?
    If Vista wasn't such a pile and didn't perform worse in games when using equivalent hardware as the same system running XP, it wouldn't be such an unappealing alternative.

    And even so, when running 4GB of RAM, how much over a LargeAddressAware flagged game with the 3GB boot.ini switch are you really gaining by using 64bit OS? Not much, really. We first need motherboards that are happy running 8GB of RAM, RAM cheap enough to buy 8GB for a reasonable price (which is not too hard with DDR2 2GB DIMMs right now), and do so at full performance/speed settings.

    Really it's not just a move to a 64bit OS, it's also a move to 8GB of RAM.

    OR it's simply having developers who code their games to work properly within 3GB of addressable RAM.
    Reply
  • instant - Saturday, July 21, 2007 - link

    How many more patches do you need for XP 64bit anyway?

    As long as the games work for it, why care about microsoft updating "security" hotfixes or not.
    Reply
  • Tegeril - Thursday, July 12, 2007 - link

    Articles like this make me very glad that I opted to go with 64bit Vista. All my hardware is supported at this point with stable drivers (we can argue the Creative X-Fi, but it works fine). I'm just amazed that people saw this coming and yet we have games that just die because of the problem. 64 bit isn't as bad as people make it out to be regarding compatibility :D - except iTunes and Quicktime :( Reply
  • halfeatenfish - Thursday, July 12, 2007 - link

    How do the *nix variants deal with this same issue? Do they even have it? Can someone shed some light, especially in terms of OS X... Leopard, if anyone knows anything there. But Tiger info is just as good. Reply
  • titan7 - Thursday, July 12, 2007 - link

    They all do the same thing for 32bit. Bu generally speaking *nix has been 64bit for years (decades) so if it is a problem just run the 64bit version of everything. And being more cross platform their code tends to have less hacks like you get in Windows apps that assume there is an extra bit available on every pointer. A single bit! Bah, we're talking about billions of bytes and elite programmers are trying to squeeze every last bit out of their application at the expense of future compatibility. LAME. Reply
  • The Boston Dangler - Thursday, July 12, 2007 - link

    OSX is a 64-bit system, *nix ymmv Reply
  • MadBoris - Thursday, July 12, 2007 - link

    Cool article Ryan. Good to see these issues getting more global attention.
    Since 32 bit seems it is here to stay for a lot longer than we want it to, and with software bloat continuing, this will hopefully continue to put pressure on driver devs to write better drivers that can handle >2GB addresses without issue. So that people can use the /3Gb switch without concern. I personally have never had problems with /3GB with any of my hardware/drivers but certainly 'less mainstream' drivers may not be handled with the care that they should be.

    I like the breakdown of games/apps that support the LargeAddressAware flag, maybe this list can grow for future articles covering more apps/games. I also enjoyed your testing on the "potential" penalty of less kernel space, something I never took the time to do on my own.

    Imagine my suprise today when making my rounds to my favorite hardware site. ;)
    Reply

Log in

Don't have an account? Sign up now