[Matthew Attoe]

Hi,

So, a Mac can now run XP natively.

BUT... how "native" would you say the set-up is?

Compared to say a HP laptop, or a Dell desktop, how do the new Macs stack up as a native PC?

I ask because I am having an issue with a MBP and a USB security dongle (please see my other thread if you have any advice or experience of this issue) and this problem doesn't exist on other Intel machines I have used.

Anyone have any info or insight?

Thanks,

Matthew

[mduell]

The only difference between an Intel Mac and the average HP/Dell machine is the Intel Mac uses EFI with a CSM to boot Windows and the HP/Dell has a BIOS. All the rest of the hardware is bog-standard.

[ghporter]

Don't post more than once on the same subject. Bad form.

mduell's statements are absolutely correct. Boot Camp's basic function is to provide a compatible bootloader for XP, and to provide specific drivers for the Mac hardware.

I do have to make one point about the "standard" part of the hardware. While all of the peripherals (video, etc.) and much of the interface hardware are indeed standard, the hardware architecture of the mainboard is different from any X86 PC. PCs have a "northbridge/southbridge" architecture to connect the CPU with the memory bus, the peripherals bus, and other system resources. The new Macs, in line with Intel's new architecture model, do NOT do this. This changes a few driver models (which is one reason Boot Camp provides drivers), and makes things a little complex for the Hardware Abstraction Layer XP uses. Not badly complex, but it's no plain-vanilla Asus motherboard we're talking about!

[meadorg]

Have you successfully used the hardware on another Core Duo platform?

[Matthew Attoe]

Hi everyone,

No, I haven;t tried it on any other Cor Duo platforms..

HOWEVER... I trashed my windows partition, reinstalled XP, installed the dongle drivers again, installed colorproof and voila, everything works a treat! Result times 10!!!!!! So, it would seem that the intel Macs are native enough for gmg software

Many thanks,

Matthew

[mduell]

Originally Posted by ghporter

I do have to make one point about the "standard" part of the hardware. While all of the peripherals (video, etc.) and much of the interface hardware are indeed standard, the hardware architecture of the mainboard is different from any X86 PC. PCs have a "northbridge/southbridge" architecture to connect the CPU with the memory bus, the peripherals bus, and other system resources. The new Macs, in line with Intel's new architecture model, do NOT do this. This changes a few driver models (which is one reason Boot Camp provides drivers), and makes things a little complex for the Hardware Abstraction Layer XP uses. Not badly complex, but it's no plain-vanilla Asus motherboard we're talking about!

You sure about that ghporter?

Intel's chipset diagram for 945PM (used in the Intel Macs) shows the usual CPU-northbridge-southbridge setup. As far as I know the only thing Apple does differently is to use EFI instead of BIOS; otherwise it should be a vanilla Asus mobo.

Do you have any links describing how the Intel Macs "hardware is radically different from a standard PC"?

[hookem2oo7]

because of intel's FSB architecture, it has to have a northbridge/southbridge system...i agree with mduell on this one

[Chuckit]

Apple says the iMac Core Duo uses a north bridge/south bridge architecture.

[mduell]

Originally Posted by hookem2oo7

because of intel's FSB architecture, it has to have a northbridge/southbridge system...i agree with mduell on this one

Even in the AMD64 architecture with the memory hanging off the CPU and a HT link from the CPU to the chipset, they have a seperate northbridge and southbridge.

[Tomchu]

Originally Posted by mduell

Even in the AMD64 architecture with the memory hanging off the CPU and a HT link from the CPU to the chipset, they have a seperate northbridge and southbridge.

Not necessarily. Many Athlon 64 chipsets are now single-chip solutions -- the nForce 3/4 come to mind right away.

With the memory controller now on the CPU, there's little need to separate the two things, as the "north bridge" would only handle PCI-Express communications. It's not cost-effective, and it's more difficult to design around. All functions of a south bridge are integrated natively into the single chip.

[ghporter]

While the structure uses north and south bridges, the architecture uses them DIFFERENTLY. My earlier post didn't say it that way though, did it? Oops.

Primarily there's the issue of PCIx versus PCI (one driver issue), and the distribution of PCI and PCIx slots (a second issue). The north and south bridges in the Intel Mac have rather different responsibilities, in large part because there's a HUGE need for PCIx for video off the north bridge, but the south bridge (traditionally tasked with peripherals) also has PCIx responsibilities, including a PCI slot for firewire connectivity (which I didn't know about until just now).

It's still a DRIVER issue; you must use a different model for drivers when there's PCIx on both bridges...

And yes, I'm a bit over my head discussing any details of the Intel Macs. Darn low pay!

[Tomchu]

There's nothing out-of-the-ordinary about the 945PM's north/south bridge combination. Take a look: http://www.intel.com/products/chipse...pm_diagram.htm

With a two-chip PCI-Express chipset, the PCI-E express always rides directly off the north bridge. If it's an Intel architecture (which it is), then the memory controller also rides off the north.

The interconnect between the two bridges is irrelevant, as the drivers will never touch anything on that low a level. In the case of some Athlon 64 chipsets, the interconnect is merely HyperTransport. In the case of this chipset, it's Intel's proprietary DMI interconnect.

The south bridge then handles everything else, and it has some PCI-Express lanes of its own so that certain devices can ride off it (WiFi, ExpressCard, GigE, etc.). The rest (like USB, SATA, PATA, audio, etc.) are natively integrated into the south bridge, avoiding the use of a peripheral bus (like PCI-E). The south bridge also provides a PCI bus, which is being utilized for Firewire on the MacBook Pro's motherboard.

This is basically a cut-and-dried two-chip PCI-Express chipset. There's absolutely nothing special or different about it. It was exactly like this in the integrated AGP days as well -- the AGP tunnel was moved upstairs, along with PCI, the integrated graphics rode off the AGP, and the south bridge handled everything else.

[mduell]

Originally Posted by Tomchu

Not necessarily. Many Athlon 64 chipsets are now single-chip solutions -- the nForce 3/4 come to mind right away.

True; I should have used the word "can" since the older chipsets seperated the two while the newer chipsets integrate the two.

Originally Posted by ghporter

While the structure uses north and south bridges, the architecture uses them DIFFERENTLY. My earlier post didn't say it that way though, did it? Oops.

Primarily there's the issue of PCIx versus PCI (one driver issue), and the distribution of PCI and PCIx slots (a second issue). The north and south bridges in the Intel Mac have rather different responsibilities, in large part because there's a HUGE need for PCIx for video off the north bridge, but the south bridge (traditionally tasked with peripherals) also has PCIx responsibilities, including a PCI slot for firewire connectivity (which I didn't know about until just now).

It's still a DRIVER issue; you must use a different model for drivers when there's PCIx on both bridges...

I'm still not seeing any significant difference. The 16x PCIe slot hanging off the northbridge takes the place of the AGP slot. The rest of the PCIe slots hang off the southbridge, just like the PCI slots used to. But this whole discussion of where the PCI/PCIe slots fall, and anything to do with Firewire, seems like a distraction. The USB ports still hang off the southbridge, as they always have; they're not on a PCI(e) bus.

Either you're way over my head in terms of depth, or one of us is flat out confused.

[ghporter]

Taking the place of the AGP slot is part of the issue; there's a different model for AGP drivers. And again, having PCIx on both the bridges means a different model still for that video driver. It's different, not radical or bizarre.

And I'm not on extremely firm ground with my understanding beyond this point. Driver models are some serious black magic, no matter what platform you're talking about, and I'm fresh out of chicken guts... Models provide a framework for communication between the OS and the device, and typically provide some hardware virtualization API. Because the connections are so different, everybody's PCIx video drivers are going to present new challenges. But beyond that, XP is built around a specific hardware philosophy, one that's very different from OS X (even on Intel hardware). They use the hardware differently, if in no other way than the methods used to access and command that hardware. I do not think XP is really "ready for prime time" in all aspects of the PCIx world, even on new PC motherboards.

[Tomchu]

Sorry to say ... but you really are quite off.

There's absolutely nothing different the 945PM chipset in comparison to any other PCI-Express chipset. The "driver model" doesn't change either. Since the integrated video is riding off the 16 PCI-E lanes that are attached to the north bridge, I really don't see why or how the video or its drivers would care what the south bridge is dealing with. You seem to be confusing extremely low-level things like chipset interconnects and architecture with drivers. Drivers are on a higher level than all of this stuff, and they really don't care how the two chipsets are connected.

You can boot up any old-style x86 operating system that isn't aware of PCI-Express and it will work just fine. You might not be able to install video drivers on such an OS, but that changes relatively little in terms of the argument. The technology is not so radical that it prevents non-PCI-E era OSes from working -- I believe there are bridges between the two buses.

OS X doesn't "use" the hardware differently than Windows. It's not like OS X is special in this regard. It's basically an x86 OS running on a relatively standard Intel x86 PC. It's not magical. :-P

[Chuckit]

Wasn't his entire point that drivers have to be done differently to deal with PCI-Express? You seem to be agreeing with him when you say, "You might not be able to install video drivers on such an OS."

[Tomchu]

Originally Posted by Chuckit

Wasn't his entire point that drivers have to be done differently to deal with PCI-Express? You seem to be agreeing with him when you say, "You might not be able to install video drivers on such an OS."

No, his entire point was that the 945PM chipset does something different, and that the "driver model" has to be changed due to how they've hooked up the different components. All I'm saying is that you might not be able to install a video driver for a PCI Express video card if the OS you're running is not aware of PCI-Express.

Windows, OS X, 945PM, south bridge/north bridge, and PCI-Express vs. AGP difference have zilch to do with anything. :-P

[mduell]

Originally Posted by ghporter

And I'm not on extremely firm ground with my understanding beyond this point. Driver models are some serious black magic, no matter what platform you're talking about, and I'm fresh out of chicken guts... Models provide a framework for communication between the OS and the device, and typically provide some hardware virtualization API. Because the connections are so different, everybody's PCIx video drivers are going to present new challenges. But beyond that, XP is built around a specific hardware philosophy, one that's very different from OS X (even on Intel hardware). They use the hardware differently, if in no other way than the methods used to access and command that hardware. I do not think XP is really "ready for prime time" in all aspects of the PCIx world, even on new PC motherboards.

This sounds like complete nonsense to me. Windows works just the same on 945PM as it does on any other PCIe or AGP chipset. The connections, structure, or architecture (whichever one you want to call it) doesn't differ from any other Intel platform.

[jwclark]

Originally Posted by ghporter

PCIx versus PCI

In case you don't know PCIx and PCI are the same thing, I think you meant to say PCIe.

[Tomchu]

Originally Posted by cavenba

In case you don't know PCIx and PCI are the same thing, I think you meant to say PCIe.

No. PCI-X is a backwards-compatible extension to PCI, but it is definitely not the "same thing".

[ghporter]

First, I had been using "PCIx" to refer to "PCI Express," completely screwing up the standard term for PCI Express. Sorry. I will use the correct term from now on.

PCI Express video memory addressing is different from AGP addressing; AGP has some particular specifics with dealing with the "memory window" that PCIe doesn't have. Remember that PCIe is supposed to be an advance beyond AGP? Part of that is how the hardware gets the data it manipulates and the instructions on how to manipulate it.

Here's what Microsoft's PCIe FAQ says about PCIe versus AGP:

PCIe Graphics & AGP

In addition to the bandwidth considerations mentioned above, there are several other differences between AGP and PCIe.

By definition, AGP requires a chipset with a graphics address relocation table (GART), which provides a linear view of nonlinear system memory to the graphics device. PCIe, however, requires that the memory linearization hardware exist on the graphics device itself instead of on the chipset. Consequently, driver support for memory linearization in PCIe must exist in the video driver, instead of as an AGP-style separate GART miniport driver. Graphics hardware vendors who want to use nonlocal video memory in their Windows XP driver model (XPDM) drivers must implement both memory linearization hardware and the corresponding software. All PCIe graphics adapters that are compatible with the WDDM must support memory linearization in hardware and software.

AGP was dedicated to graphics adapters, and no other device class used it. PCIe is intended to be used by all device classes that previously used PCI. With AGP, a number of video drivers were directly programming the chipset, which gave rise to severe ill effects such as crashing and memory corruption in the graphics stack. Because PCIe will be used for all devices in the system, it is even more important that video drivers not program the chipset directly.

In this case, "driver model" refers not only to things like addressing but limitations on what may and may not be done by the driver. THIS IS A BIG DIFFERENCE. How many PC games specify the video card chipset? Dozens! Because they demand specific commands be available - even if these commands are handled through DirectX.

The other main difference is how many lanes are available on a north bridge PCIe connection versus how many are available on a south bridge. This is literally a traffic issue, because you put a lot of stuff on the south bridge, but you put a few, very heavy users on the north bridge. These differences require differences in how the chipset handles I/O calls and THAT is often (if not always) tied up in what the device driver must/can do.

Isn't this a bit off topic for the original question? How "native" is XP on a Mac? About as native as on any brand new PC hardware. But like those fresh-off-the-press PCs, you may have to deal with new (and not thoroughly proven) drivers for some components, if those drivers are available at all. I should have just said that in the first place and not explained why I thought that way.

[Tomchu]

We understand that AGP and PCIe are completely different technologies that require different software implementations to be used to their fullest potential ...

But there is absolutely nothing different about the 945PM and its north/south bridges vs. any other two-chip PCI-Express chipset. The Mac doesn't make use of this chipset any differently than Windows.

[ghporter]

The drivers MUST be different from older, non-PCI Express chipset drivers (EVERYTHING in Windows has a driver, even the system clock and interface bus). With that said, [I]not all hardware out there that can fit will work as well as it might because of the difference in the connection between the OS and the hardware, and not all Windows drivers for PCIe hardware will be top quality for some time, if ever.

THAT IS COMPLETELY AND ENTIRELY ALL I HAD ANY INTENTION OF SAYING. I AM NOT GOING TO TRY TO CLARIFY THIS POINT EVER AGAIN. CAN WE DROP THIS LINE OF DISCUSSION?