Can't enable APIC during XP install

[Krugger]

I'm trying to reinstall XP, and i want to enable APIC. I've checked the bios, set the IO APIC to enabled, and also the ACPI O/S to enabled. Everything seems fine, but when i get to the xp installation part where you hit F5, and are supposed to see th option to select either "Standard PC" or "Other (ACPI) PC". but, i only see the standard pc choice, and other. if i select other, it asks for a disk...
any ideas as to what i've missed?

 

[JW310]

Have you just tried letting Windows determine if your motherboard supports ACPI or not? As far as I know, if Windows Setup detects that the motherboard is ACPI-compliant, then ACPI support will be enabled by default. There should be no need to select an additional driver for it.

JW

 

[Krugger]

Originally posted by: JW310
Have you just tried letting Windows determine if your motherboard supports ACPI or not? As far as I know, if Windows Setup detects that the motherboard is ACPI-compliant, then ACPI support will be enabled by default. There should be no need to select an additional driver for it.

JW

yea, i just gave up and let windows install normally, and it did enable APIC.

 

[Sianath]

Let's make sure we are getting our terms correct.

APIC = Advanced Programmable Interrupt Controller

- This is used primarily by multiproc systems, and in a uniproc system, it emulates a normal PIC. It deals with how hardware interrupts are processed.

ACPI - Advanced Configuration and Power Interface

- This deals with power management.

 

[cleverhandle]

Originally posted by: Sianath
APIC = Advanced Programmable Interrupt Controller

- This is used primarily by multiproc systems, and in a uniproc system, it emulates a normal PIC. It deals with how hardware interrupts are processed.

Hmm... this is not really my field of knowledge, but there are definitely uniprocessor APIC systems out there that do more than just emulate a traditional interrupt controller (if that's what you were describing?). Thus, with support from the OS, you can get more interrupts (24 I think?). That appears to be what the OP is looking for. But yes, APIC is definitely different than ACPI.

And yes, I would just let Windows choose the HAL on its own. It should automatically select the most advanced HAL it can use on the hardware, and forcing it to something else doesn't sound like a great idea to me.

 

[VirtualLarry]

Yes, if the chipset supports an IOAPIC, and the CPU supports APIC, and the BIOS supports the MPS spec, then the system can be switched from "legacy 8259 PIC emulation mode", to either a half-emulated/half-native mode, in which the APIC adds 8 IRQs to the existing 16, to a fully-native, non-legacy-compatible mode, that allows nearly as many IRQs as you want. Unfortunately, on most mobos, the PCI IRQ routing is done using physical "wire-OR"s, which means more than just logical sharing of IRQ numbers at the PIC/APIC.

For example, my MSI KT4V-L, support IOAPIC and MPS spec revision 1.4, and I installed W2K, XP, etc., and it auto-detects the MPS Uniprocessor HAL. (I have ACPI disabled, because it's mostly crap, and I always disable all power-management features because they never do anything but cause problems, in the long run.) When the BIOS boots up, it shows only the legacy IRQs (many shared), but after W2K boots, it shows as seperate IRQs for everything, with the AGP getting IRQ 128, the legacy/ISA IRQs getting 0-15, onboard devices getting 64-70, and PCI cards getting 20-23, 24-27, etc., basically 4 logical IRQs for each PCI slot, matching PCI INTA-INTD for each slot. Unfortunately, there is still IRQ contention, because of the physical interrupt-line wire-OR sharing, unfortunately.

When I boot Win98se (rarely), it doesn't support APICs, so it is always using the shared legacy IRQ routing. Interestingly enough, I *have* to use APIC mode in W2K, because although Via provides an IRQ routing miniport for Win98x (required for allowing proper PCI IRQ sharing), they don't provide one for W2K, claiming that one is not needed. Yet, if you install W2K on this board with the "Standard PC HAL", it clearly shows an error due to not having a driver for PCI IRQ routing. Via is full of it. I guess that they are just lazy, or want to force people to use ACPI mode or something instead. (Yes, I meant ACPI there, not APIC - MS has been trying to force ACPI upon everyone, by decreeing that mobos would not qualify/pass WHQL/PC'XXXX testing unless the "ACPI disable" setting in the BIOS was hidden or removed entirely. Most BIOS makers responded by simply making an option with a different name that does the same thing. ACPI is basically the worst thing in the world, because it has the tendency to cause all interruptss to be overlapped onto the same IRQ, which causes horrid hardware-latency problems. Yet it is the shipping default for the vast majority of systems. So sad, what the MS monopoly has done.)

 

[MrChad]

Originally posted by: VirtualLarry
Yes, I meant ACPI there, not APIC - MS has been trying to force ACPI upon everyone, by decreeing that mobos would not qualify/pass WHQL/PC'XXXX testing unless the "ACPI disable" setting in the BIOS was hidden or removed entirely. Most BIOS makers responded by simply making an option with a different name that does the same thing. ACPI is basically the worst thing in the world, because it has the tendency to cause all interruptss to be overlapped onto the same IRQ, which causes horrid hardware-latency problems. Yet it is the shipping default for the vast majority of systems. So sad, what the MS monopoly has done.

What in the world are you talking about? The IRQ sharing functionality that is part of ACPI is well within the bounds of the PCI spec, and any "horrid hardware latency problems" are only due to poorly written drivers that don't play nice with other devices. Why would you actually prefer that we return to the days of ISA and manual IRQ setting? It was and still is a nightmare to configure, and it's simply impractical nowadays given the sheer number of devices that users connect to their system.

 

[Smilin]

Originally posted by: VirtualLarry
Yes, if the chipset supports an IOAPIC, and the CPU supports APIC, and the BIOS supports the MPS spec, then the system can be switched from "legacy 8259 PIC emulation mode", to either a half-emulated/half-native mode, in which the APIC adds 8 IRQs to the existing 16, to a fully-native, non-legacy-compatible mode, that allows nearly as many IRQs as you want. Unfortunately, on most mobos, the PCI IRQ routing is done using physical "wire-OR"s, which means more than just logical sharing of IRQ numbers at the PIC/APIC.

For example, my MSI KT4V-L, support IOAPIC and MPS spec revision 1.4, and I installed W2K, XP, etc., and it auto-detects the MPS Uniprocessor HAL. (I have ACPI disabled, because it's mostly crap, and I always disable all power-management features because they never do anything but cause problems, in the long run.) When the BIOS boots up, it shows only the legacy IRQs (many shared), but after W2K boots, it shows as seperate IRQs for everything, with the AGP getting IRQ 128, the legacy/ISA IRQs getting 0-15, onboard devices getting 64-70, and PCI cards getting 20-23, 24-27, etc., basically 4 logical IRQs for each PCI slot, matching PCI INTA-INTD for each slot. Unfortunately, there is still IRQ contention, because of the physical interrupt-line wire-OR sharing, unfortunately.

When I boot Win98se (rarely), it doesn't support APICs, so it is always using the shared legacy IRQ routing. Interestingly enough, I *have* to use APIC mode in W2K, because although Via provides an IRQ routing miniport for Win98x (required for allowing proper PCI IRQ sharing), they don't provide one for W2K, claiming that one is not needed. Yet, if you install W2K on this board with the "Standard PC HAL", it clearly shows an error due to not having a driver for PCI IRQ routing. Via is full of it. I guess that they are just lazy, or want to force people to use ACPI mode or something instead. (Yes, I meant ACPI there, not APIC - MS has been trying to force ACPI upon everyone, by decreeing that mobos would not qualify/pass WHQL/PC'XXXX testing unless the "ACPI disable" setting in the BIOS was hidden or removed entirely. Most BIOS makers responded by simply making an option with a different name that does the same thing. ACPI is basically the worst thing in the world, because it has the tendency to cause all interruptss to be overlapped onto the same IRQ, which causes horrid hardware-latency problems. Yet it is the shipping default for the vast majority of systems. So sad, what the MS monopoly has done.)

:roll:

 

[LiLithTecH]

Originally posted by: VirtualLarry
Yes, if the chipset supports an IOAPIC, and the CPU supports APIC, and the BIOS supports......

There's 2 minutes (had to reread it twice) of my life I'll never get back.

 

[VirtualLarry]

Originally posted by: MrChad
What in the world are you talking about? The IRQ sharing functionality that is part of ACPI is well within the bounds of the PCI spec, and any "horrid hardware latency problems" are only due to poorly written drivers that don't play nice with other devices. Why would you actually prefer that we return to the days of ISA and manual IRQ setting? It was and still is a nightmare to configure, and it's simply impractical nowadays given the sheer number of devices that users connect to their system.

No, even MS themselves document the latency issues, and advocate that system developers move away from PCI IRQ sharing to native APIC configurations that support message-passing based interrupt mechanisms rather than physical wire-OR'ed configurations. Just because PCI IRQ sharing is allowed by the spec, does not make it optimal, and in cases in which IRQs are shared, even the most well-engineered device-driver will be forced to perform sub-optimally in that condition. The latency that I speak of, is even visible to the end-user, in the unfortunately more-common cases in which an ACPI configuration has forced all possible sharable hardware interrupts onto one singular IRQ.

And what does IRQ sharing have to do with manual IRQ configuration settings? The issues are orthagonal.
In fact, the "sheer number of devices" suggests even more that IRQ sharing should be abolished entirely, which is exactly what MS advocates to hardware developers, to avoid the inherent latencies that sharing IRQs causes.

 

[Auric]

Originally posted by: Krugger
any ideas as to what i've missed?

The down key?

 

[Smilin]

You guys are acting like shared IRQs under ACPI means multiple devices get one wire to the CPU or something.

Much shenanigans in this thread.

 

[VirtualLarry]

Originally posted by: Smilin

You guys are acting like shared IRQs under ACPI means multiple devices get one wire to the CPU or something.

Much shenanigans in this thread.

No, one is logical IRQ sharing, which causes additional software interrupt latency, and unnecessarily calling of the IRQ handler chain. The other is physical interrupt-line sharing, which depending on the APIC configuration mode and a few other things, may or may not result in the former as well. Most mobos are physically wire-ORed for the PCI INTA-INTD interrupt lines though.