Results! IRQ Sharing and lots of sound cards....

Ethan telmnstr at 757.org
Thu Aug 14 16:57:58 PDT 2003


A while ago I posted about an effort to put a good number of sound cards
in a "telcom" style backplane (PICMG) system. Well, here are the results..

First, I tried with 6 sound cards. The 6 cards appeared to work, but some
were sharing IRQs. I tested encoding audio (Darkice/LibLAME) and two cards
on the same IRQ worked flawlessly, so IRQ sharing is not an issue.

Next the partner in the project gave me 3 more Sound Blaster cards.
Problem was they had game ports, a big no-no. I tried going thru Creative
(developer relations) to find out if I could yank one of the surface mount
resistors or diodes to disable the game port, but they were of no help.
The Game/Midi port I believe only has 2 addresses availible. Never fear,
eBay to the rescue.

Bad news is 15 cards wouldn't work.

Good news is that the CPU should only be able to support encoding 10
streams as it is, running a PIII 600mhz with Darkice set to 0.8 for
"quality", sampling from the cards at 22.5khz mono.

Better news is 11 cards worked! Sometimes a 12th would show up with an
error saying there is a memory address conflict.

$ uname -a
FreeBSD audio10.757tech.net 5.0-RELEASE FreeBSD 5.0-RELEASE #0: Tue Feb 17
03:42:53 GMT 1998
root at audio10.757tech.net:/usr/src/sys/i386/compile/757AUDIO  i386
$ dmesg | grep pcm
pcm0: <AudioPCI ES1371-B> port 0xa400-0xa43f irq 11 at device 7.0 on pci2
pcm1: <AudioPCI ES1371-B> port 0x8000-0x803f irq 11 at device 4.0 on pci3
pcm2: <AudioPCI ES1371-B> port 0x8400-0x843f irq 15 at device 5.0 on pci3
pcm3: <AudioPCI ES1371-B> port 0x8800-0x883f irq 10 at device 6.0 on pci3
pcm4: <AudioPCI ES1371-B> port 0x8c00-0x8c3f irq 12 at device 7.0 on pci3
pcm5: <AudioPCI ES1371-B> port 0x9000-0x903f irq 11 at device 8.0 on pci3
pcm6: <AudioPCI ES1371-B> port 0xc000-0xc03f irq 12 at device 4.0 on pci4
pcm7: <AudioPCI ES1371-B> port 0xc400-0xc43f irq 11 at device 5.0 on pci4
pcm8: <AudioPCI ES1371-B> port 0xc800-0xc83f irq 15 at device 6.0 on pci4
pcm9: <AudioPCI ES1371-B> port 0xcc00-0xcc3f irq 10 at device 7.0 on pci4
pcm10: <AudioPCI ES1371-B> port 0xd000-0xd03f irq 10 at device 4.0 on pci5
pcm11: <AudioPCI ES1371-B> port 0xd400-0xd43f irq 12 at device 5.0 on pci5
$

$ dmesg | grep bridge
pcib0: <Intel 82443BX (440 BX) host to PCI bridge> at pcibus 0 on
motherboard
agp0: <Intel 82443BX (440 BX) host to PCI bridge> mem
0xe0000000-0xe3ffffff at device 0.0 on pci0
pcib1: <PCIBIOS PCI-PCI bridge> at device 1.0 on pci0
isab0: <PCI-ISA bridge> at device 7.0 on pci0
pci0: <bridge, PCI-unknown> at device 7.3 (no driver attached)
pcib2: <PCI-PCI bridge> at device 17.0 on pci0
pcib3: <PCI-PCI bridge> at device 18.0 on pci0
pcib4: <PCI-PCI bridge> at device 19.0 on pci0
pcib5: <PCI-PCI bridge> at device 20.0 on pci0


NOT BAD! PC hardware has always rated low in my opinion because of the IRQ
limitations, but how can you complain about this?????! The pci3 pci4 pci5
is due to the multiple PCI bridges. I'm assuming bus #1 is on the PICMG
single board computer and the others are from the large chips on the
backplane board (PCI spec calls for a max of 5 PCI slots).


There are pictures of the physical system here:
http://users.757.org/~ethan/pics/geek/soundcard_champion/

Notice there are 12 cards in there, but only 11 are detected. I am going
to pull 2 cards when I put all of the slot covers on and prepare the
system for installation in it's role as one of two streaming audio feeds.

Cards are Creative SoundBlaster 128 PCI CT-5907 I believe (ends in 07, no
game port!!).







More information about the freebsd-hardware mailing list