Right.
It's just like back in the 286 days. When you put in a whole megabyte of RAM, you lost the upper 384KB due to video memory, EMS frames BIOS, memory mapped network cards, etc. If you were lucky, your chipset could relocate the lost 384KB of ram to physical addresses just above 1MB. In real mode the CPU couldn't access the memory because real mode X86 is limited to 20 bits of physical address. However, add EMM386 to the config.sys and the CPU went into an extended X86 mode which allowed those 20 bits of address space to be mapped to a much larger physical address space. EMM386 could then map part of the relocated 384KB of RAM back down into unused spots of the area between 640K and 1MB. The same thing is happening here, except that the limitation is artificial instead of physical. My point of curiosity was to find out where exactly where server 2003 standard and XP are throttled. Was it at the physical address space level, or the amount of usable RAM level? The document referenced by Ken Schaefer clearly stated that server 2003 standard was limited to 4GB of physical address space. This would mean that even if the chipset could relocate the .5GB to an area above 4GB AND the PAE switch was used to get 36 bit addresses, you still would not get the additional RAM back because of the OS limit. On the other hand, if the 4GB limit was a RAM limit and not a physical address space limit AND your chipset could relocate the shadow memory AND the PAE switch was used, you should get some of the memory back. So we are now stuck on the horns of an enema, um, I mean dilemma. We have documentation that implies that the PAE switch should *never* give memory back regardless of the chipset capabilities, and we have empirical evidence from Edward Ziots that adding the PAE switch does indeed return some of the missing memory. I'm figuring that the documentation is wrong, or possibly was never updated to match a change in the code at some point. I realize there are other possible explanations for Ed's observations, but I'm going with what I consider the simplest: the docs are wrong. Besides, from an OS point of view, it makes more sense to me to simply limit the amount of RAM that is usable than to limit the physical address space which might make you incompatible with certain motherboards or PCI cards. ________________________________ From: Eric E Eskam [mailto:[EMAIL PROTECTED] Sent: Monday, January 14, 2008 9:34 AM To: NT System Admin Issues Subject: RE: HP DL380 G5 and Win2k3 R2 Standard not showing maximum memory in OS "Ken Cornetet" <[EMAIL PROTECTED]> wrote on 01/11/2008 07:07:10 AM: > I want to point out that I don't know where Microsoft's > artificial limit lay - with the logical RAM addressing or with > the physical addressing, hence me asking if someone could try > the /PAE switch to see if it worked. /PAE with XP on my laptop makes no difference - I see a little over 3.5 gig no matter what I do. As others have posted, you get varying results on different machines due to the chipset you have, how many PCI devices you have and how your BIOS is configured. This issue is why Apple didn't ship early MacBooks with more then 3 gigs of RAM - the pre-945 Intel chipset simply couldn't offer the extra memory for use (see, even Mac OS X can't change the characteristics of hardware). You are seeing artifacts of the Intel X86 architecture combined with 32bit operating systems - it's not pretty, nor consistent. Newer chipsets like the Intel 945 chipset family do not have these limitations on some hardware - on a MacBook Pro, Mac OS X can see all 4 gigs of RAM. On my Dell Latitude D630, I can only see a little over 3.5 gigs of RAM. Dell is doing something different then Apple, and I would wager it's the legacy PCI support that is consuming that half a gig or so of RAM. Legacy PCI support is obviously not an issue for Apple. Did that help? Eric Eskam =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= The contents of this message are mine personally and do not reflect any position of the U.S. Government "The human mind treats a new idea the same way the body treats a strange protein; it rejects it." - P. B. Medawar ~ Upgrade to Next Generation Antispam/Antivirus with Ninja! ~ ~ <http://www.sunbelt-software.com/SunbeltMessagingNinja.cfm> ~
