[Takaharu]

This is something I'm merely curious about more than anything.

Generally 32-bit operating systems can handle (approximately) 3.25GB RAM. However, some higher-end server 32-bit operating systems (Server 2008 Enterprise, for example) can handle more than 4GB RAM.
I was under the impression that it has something to do with the x86 architecture but if so, how comes x86 server operating systems can utilise more than 4GB RAM?

[Kizo]

No 32bit OS can address more than 2^32 physical memory addresses(exactly 4GB, of which a portion is used for I/O mapping, that is why you get cca 3.25GB for user usage).
I assume that what the Server OSes are doing is through PAE, which must also be supported by the CPU.

[IvanV]

Most consumer OS-es use a byte as the addressable unit. Since they have 32-bit memory addresses, the maximum amount of memory they can address is 2^32 bytes = 4GB.

If a different addressing mechanism is used, for example, two registers are used for addressing, one to point to a segment of memory and the other to point to a location within that segment, you can address more memory than you normally would, but for such schemes you need the CPU, the motherboard and the OS to support it.

Still, this usually means a wider than normal address bus. For example, Pentium had 32-bit registers and 32-bit address bus so it could go up to 4GB. Pentium pro got a wider address bus, 36-bit, so it could support as much as 64GB RAM. As for the later 32-bit x86 CPUs, I'm not sure. It looks like all PII, PIII and P4 CPUs had the support for the same 36-bit address bus, but it was not used in desktop CPUs (it was probably not even supported by the desktop chipsets and motherboards), only in the server CPUs, Xeons, derived from them. The server motherboards did and obviously the OSes were adjusted accordingly.

[Kizo]

As i said... PAE

Originally Posted by IvanV

Still, this usually means a wider than normal address bus. For example, Pentium had 32-bit registers and 32-bit address bus so it could go up to 4GB. Pentium pro got a wider address bus, 36-bit, so it could support as much as 64GB RAM. As for the later 32-bit x86 CPUs, I'm not sure. It looks like all PII, PIII and P4 CPUs had the support for the same 36-bit address bus, but it was not used in desktop CPUs (it was probably not even supported by the desktop chipsets and motherboards), only in the server CPUs, Xeons, derived from them. The server motherboards did and obviously the OSes were adjusted accordingly.

[IvanV]

Yeah, I started that post before you finished yours than got distracted so when I eventually posted it, I saw that you had already answered 10 minutes before.

[Tipstaff]

HERE's a great article that discusses things in way too much freaking detail.

[WxMan1]

OR, you could just read my post here and forgetaboutit.



BOTTOM LINE: IF you think either Vista or Win7 sucks, then either you don't know what I wrote, or what's contained at the linky provided by Tipstaff. Do I care? He's got more white-checkmarks in his rep than I do.

My story is that ALL problems attributed to Vista & Win7 are entirely related to user ID 10-T errors; that the story and one I'm sticking to it oot.



Virtually 2,5619% of user ID 10-T errors can be traced to ignorance (directly), but sometimes its pure stupidity (and quite infrequently outright foolhardiness). But, like bliss, ignorance is a very fragile thing (while the latter can be extremely tenacious - akin to a bad girlfriend); for the former the least iota of knowlege and ignorance is totally shattered, albeit for the latter only herpes will get rid of you. Anyways, I digress.

While ignorance may reform itself - akin to the metal motor-cycle cop terminator dude - it can NEVER manifest itself in the same magnitude as previously.