I’m sorry but “arrogant dickheads” like me design the technology which you use to comment on threads like this. So forgive me for not giving two hoots what you think.

chelgrain’s right, it would go over your head unfortunately. By taking offense you put yourself in the category of people he mentions.

Not everyone on the planet is as smart as you (most clearly) are and there is no need to take a swipe and people who may not have your massive brain and insight. So there are some inaccuracies – it would be helpful to point them out in a polite manner that encourages people to learn rather than beat them over the head with your arrogance.

So Apple Insider has got it all wrong… well Mr Brain Box… care to enlighten us all (including AI) as to the rest of the issues with the article – maybe we could all learn something from your clearly superior intellect… maybe you could learn something too… for example, how not to be an arsehole.

I could write thousands of words about the subject but the explanation would be completely inaccessible to the very people I would be trying to correct :/

However to try and correct some of the greater ills…

Firstly Windows, as of SP2, does use PAE all the time. This is because having PAE enabled is required to use NX or NoExecute bit. However all 32 bit consumer versions of Windows are limited to 4GB of physical RAM even with PAE enabled. As the article says this is due to broken kernel mode drivers which can’t deal with device memory above 4GB.

Secondly you always need some physical address space to map device memory into, there is no such thing as an Intel based PC that “doesn’t us memory mapped I/O” the only question is how much physical address space you need and where you can map it in the address space. Current Intel chipsets have 33 bits (8GB) of physical address space therefore you can put about 7GB of physical RAM in them and need about 1GB for device memory.

This is why Apple uses workstation class chipsets in the MacPro which can deal with far more physical address space.

You state Alienware’s top-of-the-line machine is subject to the memory shortcomings illustrated in the article, but according to their website, their machines do use Intel’s Core 2 processors but are opting for the NVIDIA 790i Ultra SLI Chipset instead. I couldn’t find any reference to memory mapping on the site, so is it possible they don’t use MMIO and are not subject to the same problems as other OEMs?

The reason I bring it up is because I usually take your words and construct a snarky graphical representations of them to further drive the point home. You use graphs and pie charts, I make ads.

So before I make one for this article, I need to know whether Alienware’s machines or any other OEM’s machines that use something other than Santa Rosa chipsets have the same memory shortcomings.