4 PCI SLOTTERS check this out

[Zapple]

There have been many posts regarding the KY7A and problems that arises when using 4 PCI slots. CHeck this email out from one of my gurs:
You can't use 4 PCI card with the KT7A - or any other motherboard that has
an on-board RAID controller.

The PCI specification says that there can ONLY be 4 bus-mastering PCI
devices on the bus at once.
No matter what.

The on-board RAID is a PCI device, but it doesn't take a slot.

Almost all PCI devices are bus-mastering, so you won't be able to use more
than 3 of them.

That is NOT abit's fault, but a basic design decision back in '92 when PCI
was invented.

The reason that motherboards with 5 PCI slots existed was back in the
Voodoo2 days.
The voodoo2 was a very simple device that did NOT require bus-mastering or
an irq.
Since many people had 2 voodoo2 cards, then with 5 pci slots, they could use
3 other PCI devices without a problem.

Needless to say, the people who are bitching at ABIT about this are just
ignorant as to how the basic subsystems in a computer work.
If they want to use 4 PCI cards, all they have to do is disable the onboard
RAID controller.

This phenomenon occurs on ALL motherboards from ALL manufacturers.

The only way around it is to put a dedicated PCI arbiter chip that is
separate from the motherboard chipset (basically a 2nd PCI controller)
That will allow you 4 more bus-mastering devices, since there will be in
effect, 2 PCI buses in the system.

This is quite expensive however, and you'll only see this feature on
high-end server motherboards that NEED that many devices (like Tyan,
Supermicro, etc.)

Feel free to post this message verbatim to whoever/wherever was discussing
this issue if nobody has set them straight yet.

 

[imonarock]

Well for us ignorant people who dont know the PCI Standard... Then Abit should have told us that the motherboard they had made with 6 pci slots can only be utilized without the raid controller. What a damn crock...

 

[Zapple]

I didn't say we'd be happy about it. This just expalins why.

 

[Vamp Armand]

That doesn't explain the reason why Abit's boards can't handle multiple PCI cards without sharing IRQ's.... while the same chipset motherboard from Asus can. Abit needs to stop making excuses.

My system:

ATI Radeon 64 VIVO
SB Live
Network Card
Promise Raid Card


My system using the KT7 had 2 devices sharing IRQ's... no matter what combination I did....that is with com1, com2, parallel printer, usb and secondary IDE disabled.

Same setup using a KK266 from Iwill does not share even 1 IRQ.

PCI bus mastering does not explain that.

 

[Peter]

Sorry, wrong. All wrong.

The PCI specification only has four separate interrupt lines, meaning that PCI devices have to be able to share interrupt. The specification has that mandatory, there is no way around this as soon as you have more than four PCI functions on a mainboard. So there is just no point in complaining about this - except if some poorly written driver screws it up.

BUT: This is nothing to do with bus master devices. There is NO limitation as to the number of bus mastering devices. For a bus mastering PCI device, you need a pair of bus request and grant signals. How many of those you got depends on the system chipset (to be precise, on the chip that creates that PCI bus), but can be expanded through an extra arbiter chip.

But this is nothing to do with the interrupts! All 100 MHz chipsets do have 5+1 req/gnt signal pairs, the extra one being for the chipset south bridge - so 5-slot boards can be built without extra circuitry. AGP bus has its own req/gnt pair as well.

AMD 761 chipset even has 7+1, and if you use a bridged PCI bus, you'll find that most bridge chips drive 9 (nine!) bus masters on their downstream size.

Regards, Peter

 

[Peter]

Vamp, regarding PCI interrupt sharing ...

There's two reasons for it happening:

(a) see above, there are only four INT lines available on the PCI side of things. Now which devices and slots connect to the same one is in the mainboard's design - this is where moving cards around "helps" (not really, since again, interrupt sharing capability is a mandatory feature for any PCI and AGP device).

(b) PCI INT to system IRQ routing may have to map more than one PCI INT to a single system IRQ - this happens when you have too much non-PCI stuff enabled on the mainboard, and thus have less than four system IRQs available for PCI use. This is where disabling unused COM and LPT ports or manual IRQ routing may "improve" things.


... but this is all pointless anyway - provided your PCI devices have PCI compliant drivers as well.

Regards, Peter

 

[1921Photoelectric]

I had a similar problem with a Fic 503+ filled to the hilt. It may also have to do with plug & pay. Try manually setting your pci slots to the devices you want them to share with. Also I have had luck with changing the sequence on how irq?s are assigned but I?m not sure if your motherboard allows that. In any case do you really both serial ports all 3 ide ports and the USB cant you disable some of these? I have to ask just for sh**'s and giggles exactly what do you have in your machine that you need everything? In any case good luck.

<< I often sit back and think, &quot;I wish I'd done that,&quot; and find out later that I already have. Richard Harris on being Drunk >>

 

[ThingyNess]

Peter: I was the one who gave Zapple the info, and you're right.

I swear that I read that it was a fundamental part of the PCI spec that there were 4 REQ/GNT lines per device, but I just checked the KT133A datasheet, and it does have 5 REQ/GNT lines, plus the one PREQ/PGNT pair for the south bridge.

Having said that, generally when you have problems with many PCI cards and irq sharing, it's a simple bios issue - my MSI 6163 PRO had a bad habit of shoving every device onto one IRQ, which caused some problems with some cheapo PCI cards.

I'm currently using my ABIT BX6-2 for that reason. Same chipset, but different bios, and it happily allows me to manually steer the individual IRQ lines as I see fit. The MSI one didn't.

 

[Demon-Xanth]

Not all PCI cards are bus mastering. Many netcard and PCI modems aren't bus mastering and will have no problem living in a system with 4 bus mastering cards.

 

[Zapple]

Hey T'Ness
Glad to see you'round. Now you can see why I am in a tizzy and I don't even have the goods yet!
The &quot;PROS&quot; over at ABIT say &quot;PCI no issue&quot; &quot;You use 5 device, it work well&quot;. This coming from Bernanrd the same guy who tells me the KT7A RAID is an apporved board. Not that I have to have an apporved board but just a point to show you these guy's don't give a flying f@#*@ about giving out misleading information and one can only assume that they could careless about customer satisfaction.

 

[Vamp Armand]

I had 3 PCI slots taken. No matter what I did two of those PCI slots shared IRQ's with either the usb, IDE controller or AGP video. Look. I friggin switched motherboards and my problems went bye bye. No IRQ conflicts. Call it what you will..... but it did the same thing on 3 Abit boards... and did not do it on 3 Asus boards, an MSI board, and then the Iwill that I currently have. You stick with what you like and i'll stay away from Abit. I am not personally attacking your machine. If you got it to work then you are probably a better builder than I am. I just got tired of swapping slots... my fingers were getting burned (a joke).

 

[Shukla]

Vamp Armand:

I've had great luck with Abit boards. But.... my KT7 didn't like a NIC and my winmodem. No matter what slot, what IRQ, what hell it put me through. Finally found out it had to do with the onboard ACPI feature. I think Asus has it disabled or a better implimentation of it.

I had to hack my KT7 bios so that I could &quot;disable&quot; ACPI. Once I did that, I had to let windows reinstall all the mobo devices (usb, et. al). Now the system runs perfectly. I would have preferred not to hack the bios but..... Oh well.

So, for those out there with a KT7 and going through hell, hack the bios with modbin. Disable the ACPI and press on. Otherwise, it's a great board. No stability problems at all.

Disabling ACPI does seem to make my processor run hotter though....

 

[ThingyNess]

Vamp Armand: funny you mention that.

As I said in an earlier post, the only time I've ever had that problem was on an MSI 6163 PRO, and the current motherboard i'm using doesn't have that problem and happens to be an ABIT

The problem has to do more with an amalgam of Win2k/ACPI/bios/PnP issues than anything.

The physical motherboard itself is almost always not to blame.

- Thingy