BSD/OS 4.0 SMP Motherboard Compatibility

This page describes motherboards and/or chipsets that have been tested with BSD/OS 4.0 SMP. Some motherboards or chipsets require special treatment to run in MP mode correctly. In some cases the workaround needed to run a particular configuration requires lines to be added to the /etc/mp.config file.

Contact BSDI support if you need assistance determining what type of motherboard or chipset you have.

Configuration Comments
Tyan S1562D (Pentium)
aka: Tomcat II
Intel Triton II chipset
chips: 82371FB or SB, 82093AA
Works with some caveats:
  1. Many motherboards were shipped with defective COAST (cache) modules, these tend to work fine when the system is lightly loaded but when a second CPU is started data gets corrupted. To determine if this is your problem, remove the COAST module from the motherboard (physically), it is not sufficient to disable the 2nd level cache in the BIOS.
  2. Some motherboards are very sensitive to the placement of cards on the ISA bus (which slots particular peripherals are plugged into); shuffling ISA cards around sometimes helps matters.
  3. Some motherboards will not reliably run at the CPU's rated speed in a dual processor configuration; sometimes lowering the CPU speed helps this situation (ie: if you have a P200 attempt to run it at 166Mhz).
  4. The 2nd IDE port built into the motherboard will not work on some machines (depends on BIOS revision, newer ones are broken) with a single ATAPI CDROM drive (the drive is not detected at boot time). If an IDE hard drive is attached as master on the second port, both it and the slave ATAPI CDROM will operate in uniprocessor mode.
  5. The 2nd IDE port built into the motherboard will not work on any machine tested in multiprocessor mode; Tyan ommitted a connection from the 2nd IDE port interrupt pin to the I/O APIC (which is only used in MP mode). This can be worked around by adding the following to /etc/mp.config:
         noextclk
         irq15=ext
         irq0=ext
    when running in this mode performance is degraded since interrupts from the system clock and the 2nd IDE channel are all directed to the boot processor (instead of the 'best' processor).
    Another option to solve this problem (for hardware hackers) is to manually run a jumper wire between the interrupt pin on the 2nd IDE connector (pin 31) and the IRQ15 pin on an available ISA connector (pin D06); this also fixes the problem of not seeing a lone ATAPI drive on the 2nd port at boot time.
    A final option is to disable the onboard 2nd IDE port and use an ISA add-in card.
  6. Older Award BIOSes are not Intel MPS compliant and provide incorrect interrupt routing information to the kernel; solution: flash a more up to date BIOS from the Tyan web site.
Tyan S1564D (Pentium)
aka: Tomcat IV
Intel Triton II chipset
chips: 82371SB (PIIX4), 82093AA (IO Apic), 82349HX (memory ctl)
Reported to work, most likely has same caveats regarding IDE and clock as the Tomcat II.

Issues:

  1. When running both CPU's the motherboard is *much* more sensitive to memory timing; SIMM's that worked OK in UP mode may not work correctly with both CPU's running. Turning off 'chipset special features' in the BIOS sometimes helps, as does running the CPU's slower than their rated speed.
  2. The presence of a Creative SB16 (or, mostly likely, clones) on the ISA bus will sometimes cause erratic behavior. It is not neccesary to have the sound drivers loaded to see problems. Sometimes putting the SB16 in a different slot will help. This problem has not been reported with AWE64's, but we have no reason to believe an AWE64 will work better either.
ASUS P54NP4
S1462
(Intel Neptune chipset)
Some machines based on this chipset have a bug in their PCI implementation that causes duplicate devices to probe, bpatch pci_scan_functions and pci_scan_buses to '1' if 'cpu siomode' gives an ENOSPC error.
Once running in symmetric I/O mode, some BIOS'es POST routines do not know how to reprogram the interrupt controllers (properly), so the machine hangs trying to soft-reboot; at present the only known solution is to power cycle on boot.
Tyan S1692D (Pentium II)
Aka: Tiger
chips: 82371AB (PIIX4), 82093AA
No known problems, but 2nd IDE port has not been tested, may have similar design flaw to 1562D.
Generic: PIIX3 or PIIX4 chips (marked 82371XX) Newer chips may not be recognized automatically, if you have one of these and the clock does not advance after going into symmetric I/O mode (running 'date' several times yields the same time each time) the problem may be fixed by adding:
     extclk
to /etc/mp.config.
Tyan S1662D (PPro)
chips: 82371SB (PIIX3), 82093AA
May have problem with 2nd IDE port (not tested).
Newer BIOS versions (5.01 and above) require a newer 'cpu' program than the one shipped with RC1; please use cpu.980623 or later (below). With this version of 'cpu' the BIOS may be set for MPS 1.1 or MPS 1.4 mode.
Intel DK440LX (PII)
chips: 440LX chipset, PIIX4
No problems (2nd IDE port even works)
Bootstrap complains that extended MP config area checksum fails, this does not seem to affect functionality or performance.
Running with BIOS set for MPS 1.4 (as opposed to 1.1) compatiblity.
Supermicro P6DBS (PII)
chips: 440BX chipset, PIIX4
ISA interrupts which were shared with PCI interrupts in PIC mode trigger continuously after switching to symmetric I/O mode. The polarity of these pins must be set to 'low'. An option will be added to 'cpu' to switch all previous ISA interrupts interrupt polarities with one override, but for now the workaround is an override for each IRQ that was previously mapped to the PCI bus:
	irqN=2,N,level,low
Where N is the IRQ number in question; each IRQ that connects to a PCI device (in PIC mode) may need an override such as this.
DEC Celebris XL 5100DP (Pentium)
chips: Neptune: 82379AB, 82433NX, 83434NX
Onboard: NCR 810 SCSI, usual multi-I/O
The 'cpu' program in beta1 would dump core on this motherboard, its BIOS reports conflicting MP configuration data (which is ignored in current versions of 'cpu').
Beta1 kernels would report CPU clock speed erratically, this seems to be due to a problem with the motherboard PIT chip emulation (worked around in s/w).
ASUS P2L97 The 'cpu' program in rc1 will not bring up this board; use cpu.980623 (below) or later.
Dell PowerEdge 2300 The 'cpu' program in the 4.0 release (or cpu.980623, previously availalable on this page) will not bring up this board due to the way the BIOS desribes PCI interrupt routing on empty PCI slots. The latest version of the cpu command (below) works on this system without any overrides.

There are some additional notes regarding SCSI support on this box:
The AIC7860 (ultra-narrow) on the motherboard runs with 4.0 drivers.
The AIC7890 (ultra2-wide) isn't currently supported. The driver for this device will be available sometime in the future.
The AMI Mega-Raid controller on this board is not supported by drivers in the 4.0 release. The driver for this device will be available sometime in the future.

Watch this space for the latest status.

Dell PowerEdge 4200 No known issues with two CPU's.

If you are having problems, please try the latest version of the cpu command.

Date File
99/03/01 (4.0.1 version) cpu

Product inquiries:
     <info@bsdi.com>
Web site feedback:
     <webmaster@bsdi.com>
Copyright © 1998 BSDI, Inc.